Antibody engineering Part 5 - unige.it · of antibody and antibody fragments including bacteria, yeast, plants, insect cells, and mammalian cells • Each has advantages, potential

Antibody engineering - Part 5Antibody expression and clinical application

Il presente materiale didattico e ciascuna sua componente sono protetti dalle leggi sul copyright, sono qui proposti in forma aggregata per soli fini di studio e per uso personale.

Sono vietati forme e modi di diffusione, gratuite od onerose, diverse da quelle stabilite dal compilatore.

Antibody expressionKipriyanov, S M, Little, M, Generation of recombinant antibodies Mol. Biotechnol. 1999 12(2):173-201

• Several expression systems are available for the production of antibody and antibody fragments including bacteria, yeast, plants, insect cells, and mammalian cells

• Each has advantages, potential applications, and limitations

Kipriyanov, S M, Little, M, Generation of recombinant antibodies Mol. Biotechnol. 1999 12(2):173-201

• Bacterial expression

• Expression in eucariotic microorganisms (S. Cerevisiae)

• Insect cell expression (Baculovirus vectors)

• Plant expression

• Mammalian cell expression



Bacterial expression

Bacteria cannot assemble whole glycosylated antibodies but they are very suitable hosts for the production of antibody fragments


Yeast expression

Complete antibodies have been expressed successfully in yeast, but they contain high-mannose, multiple-branched oligosaccharides and were shown to be defective in effector functions such as complement-mediated lysis


Insect cell expression

Antibodies produced in insect cells via baculovirus vectors also contain carbohydrate structures very different from those produced by mammalian cells


Mammalian cell expression

Intact, effector function-competent antibodies have been successfully expressed in myeloma cells and also in nonlymphoid mammalian cells, which possess the mechanisms required for correct immunoglobulin assembly, posttranslational modification, and secretion


• Secreted antibody production

• Intracellular expression of antibody fragments

• Surface expression of antibody fragments

Mammalian cell expression


Intracellular expression of antibody fragments

• Eukaryotic cells can produce antibodies that function intracellularly

• Such intracellular antibodies (intrabodies) represent a powerful and promising approach to modulate the function of selected intracellular gene products in higher organisms (phenotypic knock-out) or to protect the cell from infectious agents (intracellular immunization)

Antibody constructs in cancer therapyBeckman RA, Weiner LM, Davis HM. Antibody constructs in cancer therapy: protein engineering strategies to improve exposure in solid tumors. Cancer. 2007;109(2):170–179.

in elegant experiments using human xeno-grafts that loss of the activation Fc receptorresults in reduced therapeutic effects fromrituximab and trastuzumab against lym-phoma and breast cancer models, respec-tively29. Conversely, the effects of these anti-bodies can be increased strikingly if theinhibitory Fc receptor, Fc!RIIB, is disruptedby genetic knockout. These results are sig-nificant for two reasons: first, because theyunderline the central role of Fc receptor-expressing effectors in mAb-mediated tu-mour therapy; and second, because theyindicate an important potential role formacrophages. Inhibitory Fc!RIIB is ex-pressed on macrophages but not NK cells, sothat the enhanced effect in Fc!RIIB"/" micemust be attributed to cells other than NKcells. In the clinical setting, much less infor-mation is available regarding the importanceof mAb targeting. Campath1H, which is currently involved in a Biological LicenceApplication for the treatment of chroniclymphatic leukaemia in the USA, is a potentrecruiter of effector cells in vitro and is generally felt to operate via this mechanismin vivo30. Careful studies in patients haveshown that the isotype of Campath 1H iscrucial to its therapeutic success, and that itsinteraction with host effectors might well relate to the strong surface expression andthe stability of the target antigen, CD52(Ref. 31).

SignallingWith more successful anti-cancer mAbs,such as anti-Id, rituximab (anti-CD20) andtrastuzumab (anti-HER2/neu), there is agrowing feeling that while they are able torecruit effectors, they might also employ di-rect signalling mechanisms to achieve theircytotoxic effect4. The first clinical evidencefor such activity came from anti-Id studies,which showed a strong positive correlationbetween the therapeutic efficacy of a pa-tient’s anti-Id mAb and the ability of thatmAb to trigger tyrosine phosphorylation of intracellular proteins32.Such results were consistent with in vitro studies showing thatcrosslinking of the B-cell receptor (BCR) with Ab results in growtharrest and apoptosis in both normal and neoplastic B cells. A numberof mouse lymphoma models also support a role for BCR signalling33,for example BCL1 tumour cells emerging after long periods of anti-Id-induced dormancy still express surface Id, but often show

alterations in levels of vital signalling proteins such as Syk, HS-1 andLyn (Ref. 34).

The role of transmembrane signalling produced by other anti-cancer mAbs is less clear4. Both rituximab and trastuzumab carryhuman Fc regions and, in vitro at least, recruit complement andcellular effectors. However, this activity does not distinguish themfrom a range of mAb reagents that performed well during in vitro

R E V I E WI M M U N O L O G Y TO D AY

V o l . 2 1 N o . 8 4 0 7A U G U S T

Fig. 2. Mechanisms operating with therapeutic monoclonal antibodies (mAbs). A number of poten-tial mechanisms have been identified that allow mAbs to operate in vivo. Antibodies have tradition-ally been seen as glycoproteins that protect the body by blocking invasion by microbes. The currentsuccess of mAbs in the treatment of a range of diseases (Table 1) demonstrates that this blocking function is highly efficient and can operate at a number of levels. Thus, the ‘blocking’ section showstherapeutic mAbs preventing access of a growth factor, cytokine, or other soluble mediator by binding directly to the soluble factor itself or to the factor’s receptor. Similarly, mAbs can prevent cell–cellcrosstalk by blocking receptor–ligand interaction. Once microbes have invaded, then Ab can target effector systems, such as complement and Fc receptor-bearing cells. The ‘targeting’ section showsactivation of the classical pathway of complement (C1q) and recruitment of a cellular effector againsta target cell. It is these mechanisms that have become accepted as the most likely processes that oper-ate when tumour cells become coated with IgG mAbs. In vitro data suggest that such Abs work bestwhen the targets are highly expressed and tend to remain at the cell surface when bound by mAbs (i.e.,they do not internalize). The B-cell marker CD20 is an example of such a molecule and has proved anideal target to treatment of B-cell lymphoma with mAbs (Refs 22,23). Finally, the ‘signalling’ sectionshows an alternative mechanism that might operate in vivo with certain mAbs, such as anti-Id andanti-CD20. In this situation, the mAb appears to crosslink the target molecule and deliver trans-membrane signals that control cell division. It appears that crosslinking of neighbouring receptors isa prerequisite of such activity and that mAbs might become more active in this regard when hyper-crosslinked by Fc receptor-bearing effector cells24.

Immunology Today

Target cell

mAb

Fcreceptor

Soluble factor

Target-moleculeligand

Soluble factor receptorCrosslinked receptorTarget-molecule for Ab-attackTarget-molecule forreactive cell

Signal

Block

Targeting

Blocking Signalling

C1q

mAb crosslinking

Effector cell

Reactive cell(recognizingtarget cell)

Mechanisms operating with therapeutic mAbsG

lenn

ie M

J, J

ohns

on P

W. C

linica

l tria

ls of

ant

ibod

y th

erap

y. Im

mun

ol T

oday

. 200

0 Au

g.;2

1(8)

:403

–410

.

• Solid tumors differ from normal tissue with regard to vasculature, interstitial fluid pressure, cell density, tissue structure and composition, and extracellular matrix (ECM) components

• Interstitial fluid pressure is elevated uniformly throughout a tumor and drops precipitously to normal values in the tumor’s periphery or in the immediately surrounding tissue

• Macromolecules must diffuse against this pressure gradient to penetrate tumors

Beckman RA, Weiner LM, Davis HM. Antibody constructs in cancer therapy: protein engineering strategies to improve exposure in solid tumors. Cancer. 2007 Jan. 15;109(2):170–179.

Biological properties of solid tumors


1. opsonization, which triggers killing by immune cells

2. modification of innate biological processes such as growth and apoptosis

3. delivery of a cytotoxic payload such as a chemotherapy drug, catalytic toxin, radioisotope or enzime

Antibodies attack tumors by 3 general mechanisms

form a dimer, and one anchoring domain of a kinase anchorprotein. By fusing Fab fragments, bispecific, trivalent-binding complexes are produced with molecular massesof !160 kDa. Additional stability of the complex isachieved by introducing cysteine residues in the interact-ing modules to prevent dissociation [33].

Optimization strategies for tumor-targeting antibodiesIn recent years, experimental evidence has accumulated,revealing that there is a pharmacological window suitablefor tumor targeting, which we have called the ‘tumor targetzone’ (Box 1; Figure 2). Considering the defined structuralrequirements of intermediate size and multivalency thatthe antibodies need to accommodate [34], ideal tumortargeting antibodies largely comprise bivalent and triva-lent constructs (Figure 1), thus excluding monovalent and‘high-end’ multivalent antibodies. Although there are a fewtetravalent constructs that would fit within the ‘tumortarget zone’, they present molecular stability issues thatmight diminish their in vivo functionality, and are there-fore not included in this selection. Additional experimentsare needed for better characterization of these molecules.

In a detailed evaluation of well-characterized dimericand trimeric (Table 1) antibodies, it becomes clear that notonly size and valence, but also additional attributes – lowimmunogenicity, lack of associated functional effects, andserum stability, as well as a simple molecular design tofacilitate large-scale manufacturing – are important in atumor-targeting context. According to the criteriadescribed above and the information accessible in theliterature [10,18,30,35], it appears that the minibodyand the trimerbody are among the best currently availableantibody formats for tumor targeting.

Bivalent antibodies: minibody, the best-in-class fortumor targetingThe first minibodies were generated from the scFv T84.66/212 raised against carcinoembryonic antigen (CEA) [10].The scFv was fused to the human IgG1 CH3 domain, eitherby a two-amino acid linker that resulted in a noncovalent,hingeless minibody (LD minibody), or by the IgG1 hingeand a 10-amino acid linker peptide to produce a covalent,

hinged minibody (Flex minibody). Both constructs demon-strated high affinity for CEA, each with a KD of 3–5"10#8

M. The minibody concept has been extended by the gener-ation of analogous molecules with different specificities[36], and has been applied in preclinical and clinical stu-dies [18]. Inmice bearing human colon cancer xenografts, aradiolabeled anti-CEA minibody exhibited rapid tumoruptake (21.4–32.9% ID/g) [9,37,38] and a prolonged serumhalf-life compared with a scFv monovalent counterpart(0.9–22.5% ID g#1) [39]. Furthermore, high-resolutionmicroPET images of tumor xenografts were obtained withthis fragment when labeled with positron emitters [40,41].The encouraging results obtained with minibodies in diag-nostic applications suggest that they could serve as thera-peutic agents.

Trivalent antibodiesTrimerbody, three for the ‘size’ of two. . . and somethingelse?The trimerbody is a multivalent antibody comprising ascFv connected to the collagen XVIII NC1 trimerizationsubdomain through a 21-amino acid linker [35]. Recently,the trimerization subdomain of human collagen XVIII NC1was crystallized, and the structure indicated its high

Figure 2. Definition of the concept ‘tumor target zone’ based on pharmacokinetic and antigen-binding properties of monovalent and multivalent antibodies. Schematicdiagram displaying blood clearance (orange), tumor retention (blue) and tumor penetration (green) estimates of monovalent and multivalent antibodies. Small monovalentantibodies show great tumor penetration, but fast off-rates and blood clearance, resulting in insufficient tumor uptakes and poor-contrast images. By contrast, largemultivalent antibodies show high tumor retention times, but long serum half-lives and poor tumor penetration, with a similar outcome.

Box 1. The ‘tumor target zone’

Molecular targeting is of great interest for diagnosis and therapy. Itis founded on the principle governing ligand binding to a specificmacromolecule within a tissue-of-interest. Small monovalent anti-bodies show great tumor penetration, but possess fast off-rates (kd)and blood clearance, resulting in insufficient tumor uptake and poorcontrast images. By contrast, large, multivalent antibodies havehigh tumor retention and serum half-lives, but inefficiently pene-trate the tumor. As a result, both heavy and light antibodies producesimilar outcomes. In line with these concepts, we have termed the‘tumor target zone’ as the area where recombinant antibodies thatfulfill several features for optimal tumor-targeting are located. Theseproperties include intermediate size, multivalency, stability and lackof undesired functional effects. Recent biodistribution studies[34,75] indicate that bivalent antibodies, such as diabodies(55 kDa) and minibodies (80 kDa) [10], and trivalent antibodies, suchas trimerbodies (110 kDa) [30], might be best-suited for tumortargeting owing to a higher total tumor uptake and better tumor-to-blood ratios than native IgG molecules.

Review Trends in Biotechnology Vol.28 No.7

358

Schematic diagram displaying blood clearance (orange), tumor retention (blue) and tumor penetration (green) estimates of monovalent and multivalent antibodies. Small monovalent antibodies show great tumor penetration, but fast off-rates and blood clearance, large multivalent antibodies show high tumor retention times, but long serum half-lives and poor tumor penetration.

Cuesta AM, Sainz-Pastor N, Bonet J, Oliva B, Alvarez-Vallina L. Multivalent antibodies: when design surpasses evolution. Trends Biotechnol. 2010 28(7):355-362

Definition of the concept ‘tumor target zone’ based on pharmacokinetic and antigen-binding properties of monovalent and multivalent antibodies.


In designing antitumor constructs, a suitable balance must be found between properties that promote tumor penetration and those that promoten tumor retention

the optimal binding affinity balances 2 goals:

!

• sufficiently rapid diffusion to enable penetration into the core of the tumor

• sufficiently long retention to enable signaling inhibition, internalization, or other events required for a pharmacodynamic effect


• A major determinant of speed of diffusion through tumors is molecular size: scFv fragments diffuse approximately 6 times faster than IgG, due to their smaller size and other factors

• Molecular charge and shape also affect tumor distribution

• Affinity for the target antigen (Ag) is an important variable affecting tumor distribution

Ab characteristics that influence tumor distribution

(Ref. 48) and anti-BAT (Ref. 49). With these, the mAb provides block-ing or crosslinking function on cytotoxic lymphocytes or APC insuch a way that anti-tumour T-cells and inflammatory cells residingin the immune system are stimulated and expand to a level capableof eradicating metastatic disease. In most cases, the mAb binds to itstarget on T cells or APC, and mimics the natural ligand of the targetmolecule. Thus, in the case of anti-CD40 mAb, it probably mimicsthe action of the CD40L on helper T cells and, by activating CD40!

APC, promotes the efficiency with which tumour antigens areprocessed and presented to CD8 cytotoxic T cells47. Anti-CTLA4works in a different way, by blocking the inhibitory effects ofCTLA4–B7 interaction, it appears to alter the balance of T-cell co-stimulation so that the expansion-inducing effects of CD28 pre-dominate. Regardless of the initiating mechanism, the result of treatment is the rapid and, in some cases, massive expansion of CD8cytotoxic T cells that recognize and destroy the tumours, and leavethe animals partially or completely immune against tumour re-challenge. The next step will come in seeing whether similarimmune stimulation can be achieved in patients, where tumour

loads are greater and have evolved to escape immune detection, andwhere tumours might be less immunogenic.

Concluding remarksTechnologies now in place allow the production of chimeric orhuman reagents, with long half-lives, reduced chance of stimulatinganti-antibody responses and efficient interaction with natural effec-tors. In the future it is likely that we will see an expansion of humanmAbs, made either by phage technology or in human-Ab transgenicmice, to act as blocking reagents for a variety of immune and in-fectious diseases. This technology is already being licensed for aplethora of new targets. These reagents will be directed at a varietyof cytokines, chemokines or their receptors involved in controllingthe immune system, and at extracellular receptors and ligands thatcontrol cell–cell interactions.

In the case of cancer, progress might be somewhat slower with‘naked’ mAbs. Specificity appears to be the key issue, and selectingreagents that can modulate weak, existing, immune responses is anexciting way forward.

We thank Tenovus, Cardiff, the Cancer Research Campaign and the

Leukaemia Research Fund for their support, and are indebted to colleagues

for help and discussion of the manuscript.

Martin J. Glennie ([email protected]) is at the Tenovus Research Laboratory,and Peter W.M. Johnson ([email protected]) is at the CRC MedicalOncology Unit, The Cancer Sciences Division, Southampton UniversitySchool of Medicine, General Hospital, Southampton, UK SO16 6YD.

References1 Miller, R.A. (1982) Treatment of B-cell lymphoma with monoclonal

anti-idiotype antibody. New Engl. J. Med. 306, 517–5222 Dickman, S. (1998) Cancer therapy – antibodies stage a comeback in

cancer treatment. Science 280, 1196–11973 Baumgartner, J.D. and Glauser, M.P. (1993) Immunotherapy of

endotoxemia and septicemia. Immunobiology 187, 464–4774 Cragg, M.S. et al. (1999) Signaling antibodies in cancer therapy. Curr.

Opin. Immunol. 11, 541–5475 Farah, R.A. et al. (1998) The development of monoclonal antibodies for

the therapy of cancer. Crit. Rev. Eukaryotic Gene Expr. 8, 321–3566 Berard, J.L. et al. (1999) A review of interleukin-2 receptor antagonists

in solid organ transplantation. Pharmacotherapy 19, 1127–11377 Maini, R. et al. (1999) Infliximab (chimeric anti-tumour necrosis factor

alpha monoclonal antibody) versus placebo in rheumatoid arthritispatients receiving concomitant methotrexate: a randomised phase IIItrial. Lancet 354, 1932–1939

8 Sandborn, W.J. and Hanauer, S.B. (1999) Antitumor necrosis factortherapy for inflammatory bowel disease: a review of agents,pharmacology, clinical results and safety. Inflamm. Bowel Dis. 5, 119–133

9 Saez-Llorens, X. et al. (1998) Safety and pharmacokinetics of anintramuscular humanized monoclonal antibody to respiratorysyncytial virus in premature infants and infants withbronchopulmonary dysplasia. Pediatr. Infect. Dis. J. 17, 787–791

10 Juberlirer, S.J. (1999) Acute profound thrombocytopenia followingC7E3 Fab (abciximab) therapy: case reports, review of the literatureand implications for therapy. Am. J. Hematol. 61, 205–208

R E V I E WI M M U N O L O G Y TO D AY

V o l . 2 1 N o . 8 4 0 9A U G U S T

Fig. 3. Monoclonal Ab (mAb) in the potentiation of immune responses.Recent data show that mAbs, by crosslinking certain trigger molecules oncells of the immune system, can stimulate rapid, powerful, T-cell responsesto unknown tumour antigens. Such reagents can perform as either agonists,in which case they appear to stimulate antigen presenting cells (APCs) orlymphocytes [e.g., anti-CD40 (Ref. 46) and anti-CD137 (Ref. 47)], or asantagonists that block the inhibitory signals delivered by certain receptors,such as CTLA-4 (Ref. 48). Interestingly, a number of these stimulatory targets happen to belong to the tumour necrosis factor (TNF) receptorsuperfamily and function at the level of the APC, for example, anti-CD40mAb, or the responding lymphocytes, such as anti-CD137. In most casesthe stimulatory mAbs act like a surrogate ligand and, at least in a func-tional sense, mimic the crosslinking activity of the natural ligand. The greatpotential of such reagents lies not only in their ability to augment existingineffectual responses, but in their ability to leave the treated host immune totumour rechallenge. Only time will tell if similar reagents will function in humans. Abbreviations: CTL, cytotoxic T lymphocyte; Ag, antigen.

Immunology Today

MHC I

Tumour

Tumourantigen

T-cell-based response

Stimulatory mAb(e.g., a-CD40

a-RANKa-Flt3)

Stimulatory mAb(e.g., a-CD137

a-CTLA4a-BAT)

Tumour processing MHC II priming MHC I X-priming

T cell

CTL

Cytokines(e.g., IL-12)

Co-stimulation(e.g., CD28–B7)APC

mAbs in the potentiation of immune responses

Glennie MJ, Johnson PW. Clinical trials of antibody therapy. Immunol Today. 2000 Aug.;21(8):403–410.

Therapeutic mAb metricsReichert JM. Metrics for antibody therapeutics development. MAbs. 2010 Oct.;2(6):695–700 Nelson AL, Dhimolea E, Reichert JM. Development trends for human monoclonal antibody therapeutics. Nat Rev Drug Discov. 2010; 9(10):767–774

Nature Reviews | Drug Discovery

Num

ber o

f clin

ical

can

dida

tes

150

90

100

110

120

130

140

80

70

60

50

40

30

20

10

01985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007

All human mAbsAntineoplastic onlyImmunomodulatory onlyAnti-infective onlyOther indications

sponsoring a series of Phase II studies of belimumab in rheumatoid arthritis, Sjögren’s syndrome, Waldenstrom’s disease and pre-transplantation desensitization.

Ipilimumab is an immunostimulatory mAb that targets cytotoxic T lymphocyte antigen 4 (CTLA4). The candidate was derived from Medarex’s UltiMab transgenic mouse technology and is under clinical development by Bristol-Myers Squibb. A marketing application was submitted to the FDA and the European Medicines Agency in June 2010 for the use of ipilimu-mab as a second-line treatment for metastatic melanoma. Recent Phase III study results indicate that ipilimumab alone or in combination with a gp100 peptide vaccine improved the overall survival of patients with metastatic melanoma who had received previous treatment30. Ipilimumab has been evaluated in Phase II studies of non-small cell lung cancer, breast cancer and prostate cancer, as well as brain metastases. Bristol-Myers Squibb is planning Phase III studies of ipilimumab in non-small cell lung cancer and in prostate cancer.

Approval success ratesProbabilities of success (POS) values, such as cumulative approval in the United States and transition rates between clinical phases, have inherent limitations when the calcula-tions involve cohorts with high percentages of candidates in clinical study, as is the case for human mAbs. Calculated values will vary until fates for all candidates are known. POS values for human mAbs that entered clinical study after 1996 are preliminary estimates because fates of only 31% of the 131 candidates are known (7 approved, 33 terminated), and only 20 reached Phase III trials. In addition, the values are likely to be underestimates because clinical development of therapeutic mAbs takes an average of approximately 6 years, and so the human mAb candidates that entered clinical study during the past 6 years have not had sufficient time for approval.

Nevertheless, POS values are crucial for the decision-making process used by inves-tors, as well as for strategic planning by the biopharmaceutical industry, and even preliminary estimates can be useful. Based

on the current data, the cumulative approval rate for human mAbs is 17.5%, which will increase to 23% if raxibacumab, belimumab and ipilimumab are approved. Transition rates between clinical phases (which include data for candidates currently in studies) for the human mAbs were 89% for transitions between Phase I to II; 51% for transitions between Phase II to III; and 73% for transi-tions between Phase III to approval by the FDA (FIG. 3).

The cumulative approval success rate for human mAbs is slightly higher than the 15% value that we have calculated for humanized mAbs, which first entered com-mercial clinical development in 1988 (BOX 2). As economic conditions, regulatory climate and competitive landscape can change over time, we also compared POS values for human and humanized mAbs that entered clinical development in the same period (1997–2008). The cumulative approval success rate was 17.5% for the human mAb cohort (7 approvals per 40 candidates with known fates) and 9% for the humanized mAb cohort (5 approvals per 53 candidates with known fates). Transition rates for Phase I to II and Phase II to III were higher for the human mAbs, but the Phase III to approval rate was lower (FIG. 3). Final fates were known for 31% and 40% of the human and humanized mAbs developed in this period, respectively, and the rates may change as the final fates for more candidates are determined.

Clinical indicationsPrimary therapeutic indications were identified for the 131 human mAbs that entered clinical study after 1996. Overall, most mAb therapeutics, regardless of sequence source, are developed as treatments for cancer or immunological disorders1–3. This is also the case for human mAbs, with 59 (45%) studied for cancer and 36 (28%) for immunological disorders. These proportions have remained fairly constant since 1997 (FIG. 2).

Of the 59 antineoplastic mAbs, fates are known for only 13 (22%): 2 are approved products (panitumumab and ofatumumab) and 11 candidates were terminated. The cumulative approval success rate is 15% based on currently available data. Most of the 46 human antineoplastic mAbs that are in clinical studies are at the early stages of the process, with 5 in Phase III trials and 1 (ipilimumab) in regulatory review in the United States and in the European Union. Immunomodulatory human mAbs have a higher cumulative approval success rate (33%) than either the antineoplastic or the

Figure 2 | Cumulative number of human mAbs entering clinical study between 1985 and 2008. The primary therapeutic category for the development of human monoclonal antibodies (mAbs) that entered clinical study sponsored by commercial firms between 1985 and 2008 was determined. The cumulative numbers of human mAbs that entered development for antineoplastic, immuno-modulatory, anti-infective and all other indications were tabulated. These data demonstrate the rapid growth in human mAbs in clinical research generally, and the particularly high rates of develop-ment of antineo plastic and immunomodulatory human mAbs.

PERSPECT IVES

770 | OCTOBER 2010 | VOLUME 9 www.nature.com/reviews/drugdisc

© 20 Macmillan Publishers Limited. All rights reserved10

Cumulative number of human mAbs entering clinical study between 1985 and 2008

Nelson AL, Dhimolea E, Reichert JM. Development trends for human monoclonal antibody therapeutics. Nat Rev Drug Discov. 2010 Oct.;9(10):767–774.

www.landesbioscience.com mAbs 697

Approval phases for the 23 products were affected by whether the candidate received a priority or a standard review by the FDA (Table 1). An application is assigned a priority review when the can-didate might, if approved, be a safe and effective therapy where none currently exists, or provide significant improve-ment in disease treatment. If the candi-date does not meet either of these criteria, then the application is assigned a standard review. The FDA’s current performance goal for priority reviews is 6 months for a first action for 90% of the applications submitted in any given fiscal year. The first action is not necessarily an approval and 10% of the reviews might take lon-ger than 6 months even when the first

lymphocytic leukemia, breast cancer and colorectal cancer. The mean clinical phase for the nine products was 91.0 months and the range was 50.5–140.3 months.

Only three of the 23 products were approved for indications other than immu-nological diseases or cancer. Palivizumab (MedImmune) was approved for preven-tion of respiratory syncytial virus (RSV) infection, ranibizumab (Genentech) was approved for treatment of patients with neovascular age-related macular degen-eration and denosumab (Genentech) was approved for postmenopausal osteoporo-sis. Clinical phases cannot be given for the individual products because the informa-tion is proprietary.

(11%) were murine-derived, five (18%) were chimeric, 13 (46%) were humanized and seven (25%) were fully human and derived from either a transgenic mouse or phage display platform.

Clinical phases were calculated for the 23 currently marketed mAbs that were approved after 1996. The mean clinical phase for these products was 85.7 months (Table 2). The broad range (37.3–140.3 months) was likely due to the variety of indications studied, which included solid and hematological cancers, disorders of the immune system, infectious disease and eye and bone diseases. Eleven of the 23 products (48%) were first approved as treatments for immunological indica-tions, including psoriasis, asthma, Crohn disease, multiple sclerosis, rheumatoid arthritis (RA), paroxysmal nocturnal hemoglobinuria and prevention of acute kidney rejection. A common feature of these conditions is the need for immune system modulation. The mean clinical phase for the 11 immunological mAbs was 85.6 months and the range was 59.2–113.2 months. Nine of the 23 products (39%) were approved for cancer. These mAbs were first approved as treatments for non-Hodgkin lymphoma, chronic

Figure 1. Years of first clinical study and approval for therapeutic monoclonal antibodies. *Two-year moving average. FDA, United States Food and Drug Administration.

Table 2. Mean (median) clinical and FDA approval phases for 23 therapeutic monoclonal antibodies*

mAb type Clinical phase FDA approval phase

All mAbs (n = 23) 85.7 (82.5) months 12.8 (8.9) months

Immunological (n = 11) 85.6 (83.2) months 14.6 (9.1) months

Anticancer (n = 9) 91.0 (82.5) months 11.7 (8.9) months

Priority review (n = 16) 85.2 (79.1) months 9.3 (6.0) months

Standard review (n = 7) 86.9 (88.5) months 20.9 (21.8) months*Products approved between January 1997 and July 2010. FDA, US Food and Drug Administration; mAbs, monoclonal antibodies

Reichert JM. Metrics for antibody therapeutics development. MAbs. 2010 Oct.;2(6):695–700.


• A total of 28 mAb therapeutic products have been approved by the United States Food and Drug Administration (US FDA) and 4 are undergoing regulatory review

• Global sales exceeded US $1 billion in 2009 for each of at least nine mAb products

www.landesbioscience.com mAbs 697

Approval phases for the 23 products were affected by whether the candidate received a priority or a standard review by the FDA (Table 1). An application is assigned a priority review when the can-didate might, if approved, be a safe and effective therapy where none currently exists, or provide significant improve-ment in disease treatment. If the candi-date does not meet either of these criteria, then the application is assigned a standard review. The FDA’s current performance goal for priority reviews is 6 months for a first action for 90% of the applications submitted in any given fiscal year. The first action is not necessarily an approval and 10% of the reviews might take lon-ger than 6 months even when the first

lymphocytic leukemia, breast cancer and colorectal cancer. The mean clinical phase for the nine products was 91.0 months and the range was 50.5–140.3 months.

Only three of the 23 products were approved for indications other than immu-nological diseases or cancer. Palivizumab (MedImmune) was approved for preven-tion of respiratory syncytial virus (RSV) infection, ranibizumab (Genentech) was approved for treatment of patients with neovascular age-related macular degen-eration and denosumab (Genentech) was approved for postmenopausal osteoporo-sis. Clinical phases cannot be given for the individual products because the informa-tion is proprietary.

(11%) were murine-derived, five (18%) were chimeric, 13 (46%) were humanized and seven (25%) were fully human and derived from either a transgenic mouse or phage display platform.

Clinical phases were calculated for the 23 currently marketed mAbs that were approved after 1996. The mean clinical phase for these products was 85.7 months (Table 2). The broad range (37.3–140.3 months) was likely due to the variety of indications studied, which included solid and hematological cancers, disorders of the immune system, infectious disease and eye and bone diseases. Eleven of the 23 products (48%) were first approved as treatments for immunological indica-tions, including psoriasis, asthma, Crohn disease, multiple sclerosis, rheumatoid arthritis (RA), paroxysmal nocturnal hemoglobinuria and prevention of acute kidney rejection. A common feature of these conditions is the need for immune system modulation. The mean clinical phase for the 11 immunological mAbs was 85.6 months and the range was 59.2–113.2 months. Nine of the 23 products (39%) were approved for cancer. These mAbs were first approved as treatments for non-Hodgkin lymphoma, chronic

Figure 1. Years of first clinical study and approval for therapeutic monoclonal antibodies. *Two-year moving average. FDA, United States Food and Drug Administration.

Table 2. Mean (median) clinical and FDA approval phases for 23 therapeutic monoclonal antibodies*

mAb type Clinical phase FDA approval phase

All mAbs (n = 23) 85.7 (82.5) months 12.8 (8.9) months

Immunological (n = 11) 85.6 (83.2) months 14.6 (9.1) months

Anticancer (n = 9) 91.0 (82.5) months 11.7 (8.9) months

Priority review (n = 16) 85.2 (79.1) months 9.3 (6.0) months

Standard review (n = 7) 86.9 (88.5) months 20.9 (21.8) months*Products approved between January 1997 and July 2010. FDA, US Food and Drug Administration; mAbs, monoclonal antibodies


696 mAbs Volume 2 Issue 6

voluntarily withdrawn from the US market. The approved products encompass a vari-ety of molecular origins and therapeu-tic categories. Although the first mAb was approved in 1986, most (93%) were approved after 1996, which is a period during which companies transitioned from the study of murine mAbs to chime-ric, humanized and human forms. Of the 28 products, the protein sequences of three

firms, do not necessarily identify the composition of matter of their Phase 1 candidates.

FDA-Approved mAbs

Clinical phase lengths were derived from data for FDA-approved mAbs. As of July 2010, this cohort included 28 therapeutic mAbs (Fig. 1 and Table 1), but four were

as a two-year moving average). With the increased participation by major pharma-ceutical firms, the rate rose from approxi-mately 20 per year in 2002 to over 50 per year in 2008. The decrease observed in 2009 may be real and possibly a conse-quence of companies reserving resources during an economic downturn, or it may simply be an undercount because com-panies, especially major pharmaceutical

Table 1. Therapeutic monoclonal antibodies undergoing regulatory review or approved in the United States

INN (trade name) Description, target Therapeutic category Year (review status) of first FDA approval

Muromonab-CD3 (Orthoclone OKT3) Murine, IgG2α, anti-CD3 Immunological 1986* (NA)

Abciximab (ReoPro) Chimeric, IgG1κ Fab, anti-GPIIb/IIIa Hemostasis 1994 (P)

Rituximab (Rituxan) Chimeric, IgG1κ, anti-CD20 Cancer 1997 (P)

Daclizumab (Zenapax) Humanized, IgG1κ, anti-CD25 Immunological 1997* (P)

Basiliximab (Simulect) Chimeric, IgG1κ, anti-CD25 Immunological 1998 (P)

Palivizumab (Synagis) Humanized, IgG1κ, anti-RSV Anti-infective 1998# (P)

Infliximab (Remicade) Chimeric, IgG1κ, anti-TNF Immunological 1998 (P)

Trastuzumab (Herceptin) Humanized, IgG1κ, anti-HER2 Cancer 1998 (P)

Gemtuzumab ozogamicin (Mylotarg) Humanized, IgG4κ, anti-CD33; toxin Cancer 2000* # (P)

Alemtuzumab (Campath) Humanized, IgG1κ, anti-CD52 Cancer 2000 (P)

Ibritumomab tiuxetan (Zevalin) Murine, IgG1κ, anti-CD20; Y-90 Cancer 2002 (P)

Adalimumab (Humira) Human, IgG1κ, anti-TNF Immunological 2002 (S)

Omalizumab (Xolair) Humanized, IgG1κ, anti-IgE Immunological 2003# (S)

Tositumomab-I131 (Bexxar) Murine, IgG2αλ, anti-CD20; I-131 Cancer 2003 (P)

Efalizumab (Raptiva) Humanized, IgG1κ, anti-CD11a Immunological 2003* # (S)

Cetuximab (Erbitux) Chimeric, IgG1κ, anti-EGF receptor Cancer 2004 (P)

Bevacizumab (Avastin) Humanized, IgG1κ, anti-VEGF Cancer 2004 (P)

Natalizumab (Tysabri) Humanized, IgG4κ, anti-α4β1-integrins Immunological 2004 (P)

Ranibizumab (Lucentis) Humanized, IgG1κ Fab, anti-VEGF Ophthalmic 2006 (P)

Panitumumab (Vectibix) Human, IgG2κ, anti-EGF receptor Cancer 2006# (P)

Eculizumab (Soliris) Humanized, IgG2/4κ, anti-C5 Immunological 2007# (P)

Certolizumab pegol (Cimzia) Humanized, Fab’, anti-TNF; PEG Immunological 2008# (S)

Golimumab (Simponi) Human, IgG1κ, anti-TNF Immunological 2009 (S)

Canakinumab (Ilaris) Human IgG1κ, anti-IL1β Immunological 2009# (P)

Ustekinumab (Stelara) Human IgG1κ, anti-IL12/23 Immunological 2009# (S)

Ofatumumab (Arzerra) Human IgG1κ, anti-CD20 Cancer 2009# (P)

Tocilizumab (Actemra) Humanized IgG1κ, anti-IL6 receptor Immunological 2010# (S)

Denosumab (Prolia) Human IgG2, anti-RANK-L Bone disorders 2010# (S)

Motavizumab Humanized IgG1κ, anti-RSV Anti-infective Pending (S)

Raxibacumab Human IgG1, anti-B. anthrasis PA Anti-toxin Pending (P)

Belimumab Human IgG1λ, anti-BLyS Immunological Pending (P)

Ipilimumab Human IgG1κ, anti-CTLA-4 Cancer Pending (P)

Data current as of July 2010. BLyS, B lymphocyte stimulator; C5, complement-5; CD, cluster of differentiation; CTLA-4, cytotoxic T lymphocyte-associated antigen 4; EGF, epidermal growth factor; Fab, antigen-binding fragment; FDA, United States Food and Drug Administration; GP, glycoprotein; HER, human epidermal growth factor receptor; I-131, iodine-131; IL, interleukin; INN, international non-proprietary name; NA, not applicable; P, priority review; PA, protective antigen; PEG, polyethylene glycol; RANK-L, receptor activator of NF kappa B ligand; RSV, respiratory syncytial virus; S, standard review; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor; Y-90, yttrium-90. *Voluntarily withdrawn from the US market. #As of July 2010, product had not been approved for supplemental indications.

Data

cur

rent

as

of J

uly

2010

Reich

ert J

M. M

etric

s fo

r ant

ibod

y th

erap

eutic

s de

velo

pmen

t. M

Abs.

201

0; 2

(6):6

95–7

00.


Of the 28 products, the protein sequences of 3 (11%) were murine-derived, 5 (18%) were chimeric, 13 (46%) were humanized and 7 (25%) were fully human and derived from either a transgenic mouse or phage display platform.


mAb

pip

elin

e (%

)50

45

40

35

30

25

20

15

10

5

0Humanized HumanChimericMurine

30

7

13.5

9

45

39

11.5

451990–19992000–2008

Approval was based on a single Phase III trial20, as well as on smaller proof-of-concept studies21. CAPS are rare, with the number of patients with CAPS living in the United States estimated to be in the hundreds or low thousands. In October 2009, canakinumab was approved in the European Union for patients with CAPS as young as 4 years old. Canakinumab is currently in early-stage studies for the treatment of other disorders, including rheumatoid arthritis, gout and diabetes mellitus.

Ustekinumab, another UltiMab-derived product, targets the p40 subunit shared by IL-12 and IL-23. The mAb was approved by the FDA in September 2009 as a treatment for plaque psoriasis. Two Phase III studies in patients with moderate to severe plaque psoriasis have been completed, as well as a third Phase III study comparing ustekinumab with etanercept (Enbrel; Amgen/Pfizer), a fusion protein that targets TNF, in the same patient population22. In January 2009, the European Commission approved ustekinu-mab for treating moderate to severe plaque psoriasis in adults. Ustekinumab is also currently in Phase II studies as a treatment for sarcoidosis, and in Phase III studies as a treatment for palmoplantar pustular psoriasis, palmoplantar pustulosis or psoriatic arthritis.

Ofatumumab is a CD20-specific mAb generated from the UltiMab platform. It targets a CD20 epitope that is distinct from the epitope targeted by rituximab (Rituxan/MabThera; Genentech/Biogen Idec/Roche), the pioneering CD20-specific chimeric mAb. Rituximab was approved for the treatment of non-Hodgkin’s lym-phoma in 1997, and subsequently has also received regulatory approval for the treat-ment of rheumatoid arthritis and chronic lymphocytic leukaemia23,24. Ofatumumab was approved by the FDA in October 2009, and given a conditional approval by the European Commission in April 2010, for the treatment of chronic lymphocytic leu-kaemia that is refractory to the humanized mAb alemtuzumab (Campath; Genzyme) and the nucleoside analogue fludarabine. Ofatumumab is under Phase III evaluation in patients with non-Hodgkin’s lymphoma and in patients with rheumatoid arthritis25.

Denosumab, a mAb specific for receptor activator of nuclear factor-κB ligand (RANKL), was approved by the FDA in June 2010 for the treatment of postmeno-pausal osteoporosis (PMO) in women. Trials have also been conducted to support a prevention indication for PMO, as well as for the treatment and prevention of bone loss in patients undergoing hormone ablation therapy for prostate or breast cancer26. Denosumab was approved in Europe for the treatment of PMO and of bone loss in patients with prostate cancer undergoing hormone ablation therapy. It is also undergoing regu-latory review in Switzerland, Australia and Canada for one or more of these indications.

Although not yet approved, three human mAbs — raxibacumab, belimumab and ipilimumab — are currently undergoing review by the FDA. Human Genome Sciences is the sponsor of both raxibacumab and belimumab, and Bristol-Myers Squibb is sponsoring ipilimumab. Raxibacumab binds Bacillus anthracis protective antigen and has been developed as a treatment for inhalation anthrax27. Human Genome Sciences initiated the delivery of 20,000 doses of raxibacumab to the US Strategic National Stockpile for emergency use under a contract with the US Biomedical Advanced Research and Development Authority, and an additional 45,000 doses were ordered in July 2009.

Belimumab is a human mAb specific for B lymphocyte stimulator, and was identified through use of phage-display-based tech-nologies in collaboration with Cambridge Antibody Technology. GlaxoSmithKline and Human Genome Sciences submitted marketing applications to both the FDA and the European Medicines Agency in June 2010 for the use of belimumab in systemic lupus erythematosus. This submission is based primarily on clinical and biomarker improvements in two pivotal Phase III trials in systemic lupus erythematosus — BLISS-52 and BLISS-76, which collectively involved 1,684 patients with this disease globally28 — as well as favourable post-hoc analyses of Phase II studies29. If the applica-tion is successful, belimumab will be the first new therapeutic approved for systemic lupus erythematosus in 50 years. GlaxoSmithKline and Human Genome Sciences are further

Figure 1 | Percentage of four types of mAbs in clinical development during the periods 1990–1999 and 2000–2008. Monoclonal anti-bodies (mAbs) that entered clinical study spon-sored by commercial firms between 1990 and 1999 and between 2000 and 2008 were classified according to their sequence source: murine only, chimeric (murine variable regions and human constant regions), humanized (human with murine complementarity-determining regions), and human only. These data demonstrate the substantial increase in the clinical study of human mAbs in the 2000s, a trend towards reduced use of humanization and chimeric candidates, and a dramatic reduction in the number of murine mAbs in clinical development in the 2000s.

Box 2 | Analysis criteria

Since it was founded in 1976, the Tufts Center for the Study of Drug Development has collected data on the clinical study and approval of therapeutics and vaccines. Data for monoclonal antibodies (mAbs) were collected by surveying pharmaceutical and biotechnology firms, from public documents and from commercially available databases (IDdb3, IMS R&D Focus and PharmaProjects). Data were updated with all changes that were noted until June 2010.

The data set comprises a total of 147 human mAbs that entered clinical study sponsored by commercial firms between January 1985 and December 2008, 131 of which entered study between 1997 and 2008. The status of the candidates was as follows: 88 were in clinical studies and not yet approved in any country (30 in Phase I, 51 in Phase II and 7 in Phase III); 3 were under regulatory review by the US Food and Drug Administration; 7 were approved in the United States; and 49 were discontinued. Candidates in Phase I or II were assigned to Phase II, and products in Phase II or III were assigned to Phase III. The human mAb data were compared with data for humanized mAbs that entered clinical study between 1988 and 2008 (n = 167) and between 1997 and 2008 (n = 133).

Approval success calculations were based on data for candidates with known fates (market approval in the United States or discontinuation of all clinical studies). Percentage completion was defined as the percentage of candidates with a known fate in a given cohort. Clinical-phase transition probabilities were calculated as follows: the number of candidates that successfully completed a given phase was divided by the difference between the number of candidates that entered the phase and those that were still in the phase at the time of the calculation. Transitions occurring between phases of clinical studies conducted worldwide were included.

PERSPECT IVES

NATURE REVIEWS | DRUG DISCOVERY VOLUME 9 | OCTOBER 2010 | 769


Percentage of 4 types of mAbs in clinical development during the periods 1990-1999 and 2000-2008



Tran

sitio

n ra

te (%

)

100

90

80

70

60

50

40

30

20

10

0Phase III–

US approvalPhase II–IIIPhase I–II

80

89 8794

4751

60

73

86

73

86

50

Humanized, 1997–2008 (n = 133)Human, 1997–2008 (n = 131)Transgenic mouse (n = 56)Display technology (n = 35)

(KB001; developed by Kalobios/Sanofi Pasteur, and panobacumab; developed by Kenta Biotech) and methicillin-resistant Staphylococcus aureus (Aurograb; devel-oped by Novartis). Finally, one human mAb (efungumab; developed by Novartis) targets candidal heat shock protein 90, an intra-cellular antigen released during infection.

Platform technologiesWe were able to identify the platform technologies that were used to develop 103 (79%) of the 131 mAbs analysed. Despite the availability of innovative approaches to sample natural human immune responses for the creation of mAbs10, most therapeutic human mAbs in clinical study were derived from either immunization of transgenic mice expressing human antibody genes or phage-display recombinants. The first candidate molecules from both technolo-gies entered clinical development in the late 1990s, so performance differences cannot be attributed to time-dependent variables.

Use of transgenic mice expressing human immunoglobulins avoids human anti-mouse antibody responses and main-tains the technical advantages of mouse hybridomas. Of the 103 candidates from

identified platforms, 56 were produced in transgenic mice; 6 were approved for marketing (panitumumab, golimumab, canakinumab, ustekinumab, ofatumumab and denosumab), and 15 were terminated. The current cumulative approval success rate is therefore 29%, a higher rate than that currently calculated for human mAbs as a whole (17.5%). In particular, candidates derived from transgenic mouse platforms have considerably higher Phase II to III and Phase III to approval transition rates than those of the entire cohort of human mAbs (FIG. 3). Again, it is important to note that these rates will change to some extent over time as fates for more human mAb candi-dates are determined in the future. The two primary technologies for generating human mAbs from transgenic mice were first described in 1994 (REFS 33,34) and use two different engineering approaches to inactivate endogenous mouse genes and to insert exogenous human immunoglobulin genes35,36.

A total of 34 human mAbs were identified as being derived from Medarex’s HuMAb, UltiMab, TC Mouse or KM Mouse platforms. Four of these candidates have been approved (golimumab, canakinumab, ustekinumab and ofatumumab), 1 (ipilimumab) is under regulatory review, 7 were terminated and 22 are now in clinical studies (7 at Phase I, 14 at Phase II and 1 at Phase III). Although immunoglobulin isotype was not known for all the candidates, the majority of Medarex platform-derived molecules were IgG1 (at least 21 mAbs). Of the total 34 mAbs, 16 (47%) were intended for the treatment of cancer, 13 (38%) for immunological conditions and 3 (9%) were anti-infective agents.

The XenoMouse platform developed by Abgenix, which was acquired by Amgen in 2005, was used for the development of at least 18 human mAbs. Two (panitumu-mab and denosumab) are approved, nine are in clinical study (two at Phase II, five at Phase II and two at Phase III) and seven are discontinued. Most of the XenoMouse-derived candidates were IgG2 (11 mAbs), four were IgG1, one was IgG4 and two were of unknown isotype. Two-thirds of the mAbs were cancer treatments; only two (11%) were for immunological conditions and one (6%) was an anti-infective agent.

Other transgenic mouse-based platforms, such as Regeneron’s VelocImmune and XTL Biopharmaceuticals’ Trimera, have each yielded at least one early-stage candidate.

A second popular approach for producing human mAbs is the recombinant expression of human antigen-binding fragments in a bacteriophage and subsequent selection is based on desirable antigen-binding proper-ties37–39. This technology was used to create at least 35 human mAbs that entered clinical development. Unlike transgenic mouse technologies, phage-display is used by numerous companies, including MedImmune Cambridge (formerly Cambridge Antibody Technology), Dyax, MorphoSys, BioInvent and NeuTec.

One phage-display-derived mAb (adalimumab) has been approved, and two (raxibacumab and belimumab) are under review by the FDA. In addition, 3 phage-display-derived mAb candidates are in Phase I, 19 are in Phase II and 3 are in Phase III. So far, development of seven mAbs has been discontinued. The current cumulative approval success rate for all phage-display-based technologies is 12.5%, although this is

Figure 3 | Transition rates between clinical phases for human mAbs. The historical rates of transition from Phase I to II, Phase II to III and Phase III to review by the US Food and Drug Administration are depicted. The review to approval rate was 100% for all human mono-clonal antibodies (mAbs) in the categories presented. Data for human mAbs derived from transgenic mouse and display technologies are shown separately and data for humanized mAbs are included for comparison.

Glossary

AllotypeAntibody allotypes are defined by their polymorphism within the immunoglobulin heavy and light chains. Natural allelic genetic variation in the constant region of genes in humans may predispose a given patient to anti-drug antibody responses if the drug is a foreign allotype.

Ankylosing spondylitisA chronic condition of unknown aetiology that is characterized by inflammation of the joints of the spine and pelvis. Disease progression may result in fusion of the joints.

Anti-idiotypic antibodyAn antibody that targets the hypervariable antigen-binding domain of an exogenous immunoglobulin, including therapeutic monoclonal antibodies. As the constant regions are fairly conserved, with the exception of allotypic differences, many anti-immunoglobulin responses will be directed against the highly variable, antigen-binding domain.

Cryopyrin-associated periodic syndromes(CAPS). A group of rare, inherited autoimmune disorders associated with over-secretion of interleukin-1 that may cause inflammation of the skin, eyes, bones, joints and meninges.

Phage-display technologiesA method involving the use of bacteriophages to select desirable antibody variable domains based on their binding properties.

Pre-transplant desensitizationIn the recipient patient, reduction of antibody- producing cells or the amount of circulating antibodies that might target foreign tissue prior to transplantation of an organ.

Systemic lupus erythematosusA chronic, inflammatory autoimmune disease affecting connective tissue throughout the body.

PERSPECT IVES




Tran

sitio

n ra

te (%

)

100

90

80

70

60

50

40

30

20

10

0Phase III–

US approvalPhase II–IIIPhase I–II

80

89 8794

4751

60

73

86

73

86

50

Humanized, 1997–2008 (n = 133)Human, 1997–2008 (n = 131)Transgenic mouse (n = 56)Display technology (n = 35)

(KB001; developed by Kalobios/Sanofi Pasteur, and panobacumab; developed by Kenta Biotech) and methicillin-resistant Staphylococcus aureus (Aurograb; devel-oped by Novartis). Finally, one human mAb (efungumab; developed by Novartis) targets candidal heat shock protein 90, an intra-cellular antigen released during infection.

Platform technologiesWe were able to identify the platform technologies that were used to develop 103 (79%) of the 131 mAbs analysed. Despite the availability of innovative approaches to sample natural human immune responses for the creation of mAbs10, most therapeutic human mAbs in clinical study were derived from either immunization of transgenic mice expressing human antibody genes or phage-display recombinants. The first candidate molecules from both technolo-gies entered clinical development in the late 1990s, so performance differences cannot be attributed to time-dependent variables.

Use of transgenic mice expressing human immunoglobulins avoids human anti-mouse antibody responses and main-tains the technical advantages of mouse hybridomas. Of the 103 candidates from

identified platforms, 56 were produced in transgenic mice; 6 were approved for marketing (panitumumab, golimumab, canakinumab, ustekinumab, ofatumumab and denosumab), and 15 were terminated. The current cumulative approval success rate is therefore 29%, a higher rate than that currently calculated for human mAbs as a whole (17.5%). In particular, candidates derived from transgenic mouse platforms have considerably higher Phase II to III and Phase III to approval transition rates than those of the entire cohort of human mAbs (FIG. 3). Again, it is important to note that these rates will change to some extent over time as fates for more human mAb candi-dates are determined in the future. The two primary technologies for generating human mAbs from transgenic mice were first described in 1994 (REFS 33,34) and use two different engineering approaches to inactivate endogenous mouse genes and to insert exogenous human immunoglobulin genes35,36.

A total of 34 human mAbs were identified as being derived from Medarex’s HuMAb, UltiMab, TC Mouse or KM Mouse platforms. Four of these candidates have been approved (golimumab, canakinumab, ustekinumab and ofatumumab), 1 (ipilimumab) is under regulatory review, 7 were terminated and 22 are now in clinical studies (7 at Phase I, 14 at Phase II and 1 at Phase III). Although immunoglobulin isotype was not known for all the candidates, the majority of Medarex platform-derived molecules were IgG1 (at least 21 mAbs). Of the total 34 mAbs, 16 (47%) were intended for the treatment of cancer, 13 (38%) for immunological conditions and 3 (9%) were anti-infective agents.

The XenoMouse platform developed by Abgenix, which was acquired by Amgen in 2005, was used for the development of at least 18 human mAbs. Two (panitumu-mab and denosumab) are approved, nine are in clinical study (two at Phase II, five at Phase II and two at Phase III) and seven are discontinued. Most of the XenoMouse-derived candidates were IgG2 (11 mAbs), four were IgG1, one was IgG4 and two were of unknown isotype. Two-thirds of the mAbs were cancer treatments; only two (11%) were for immunological conditions and one (6%) was an anti-infective agent.

Other transgenic mouse-based platforms, such as Regeneron’s VelocImmune and XTL Biopharmaceuticals’ Trimera, have each yielded at least one early-stage candidate.

A second popular approach for producing human mAbs is the recombinant expression of human antigen-binding fragments in a bacteriophage and subsequent selection is based on desirable antigen-binding proper-ties37–39. This technology was used to create at least 35 human mAbs that entered clinical development. Unlike transgenic mouse technologies, phage-display is used by numerous companies, including MedImmune Cambridge (formerly Cambridge Antibody Technology), Dyax, MorphoSys, BioInvent and NeuTec.

One phage-display-derived mAb (adalimumab) has been approved, and two (raxibacumab and belimumab) are under review by the FDA. In addition, 3 phage-display-derived mAb candidates are in Phase I, 19 are in Phase II and 3 are in Phase III. So far, development of seven mAbs has been discontinued. The current cumulative approval success rate for all phage-display-based technologies is 12.5%, although this is

Figure 3 | Transition rates between clinical phases for human mAbs. The historical rates of transition from Phase I to II, Phase II to III and Phase III to review by the US Food and Drug Administration are depicted. The review to approval rate was 100% for all human mono-clonal antibodies (mAbs) in the categories presented. Data for human mAbs derived from transgenic mouse and display technologies are shown separately and data for humanized mAbs are included for comparison.

Glossary

AllotypeAntibody allotypes are defined by their polymorphism within the immunoglobulin heavy and light chains. Natural allelic genetic variation in the constant region of genes in humans may predispose a given patient to anti-drug antibody responses if the drug is a foreign allotype.

Ankylosing spondylitisA chronic condition of unknown aetiology that is characterized by inflammation of the joints of the spine and pelvis. Disease progression may result in fusion of the joints.

Anti-idiotypic antibodyAn antibody that targets the hypervariable antigen-binding domain of an exogenous immunoglobulin, including therapeutic monoclonal antibodies. As the constant regions are fairly conserved, with the exception of allotypic differences, many anti-immunoglobulin responses will be directed against the highly variable, antigen-binding domain.

Cryopyrin-associated periodic syndromes(CAPS). A group of rare, inherited autoimmune disorders associated with over-secretion of interleukin-1 that may cause inflammation of the skin, eyes, bones, joints and meninges.

Phage-display technologiesA method involving the use of bacteriophages to select desirable antibody variable domains based on their binding properties.

Pre-transplant desensitizationIn the recipient patient, reduction of antibody- producing cells or the amount of circulating antibodies that might target foreign tissue prior to transplantation of an organ.

Systemic lupus erythematosusA chronic, inflammatory autoimmune disease affecting connective tissue throughout the body.

PERSPECT IVES



Transition rates between clinical phases for human mAbs


Therapeutic antibody fragment metrics

332 VOLUME 27 NUMBER 4 APRIL 2009 NATURE BIOTECHNOLOGY

which were fragments. Of these, three have been approved in the United States (Table 1), one has been approved in China, 19 (35%) are currently in clinical study and 31 (57%) have been discontinued (Fig. 1). Of the 54 antibody fragments, 30 are Fabs (56%), 19 are scFvs (35%) and 5 are third-generation versions such as miniaturized antibodies (9%).

Eight bispecific fragments (15%) have entered clinical development, all as antine-oplastic agents. Two bispecific scFvs are in active clinical development (that is, under evaluation in at least one ongoing or recently completed clinical study of any type), whereas six bispecific Fab candidate fragments have been discontinued. Twenty-four fragments (44%) were conjugated to additional func-tional molecules, such as radioisotopes or cytotoxins. These immunoconjugate candi-dates were all studied as anticancer agents; eight are in current clinical study, including two candidates in phase 3.

The rate of introduction of fragments into clinical development has varied over time. Only 3 agents entered clinical study during the 1980s, but an average of 1.3, 2.6 and 3.4 candidates per year were first studied in the 1990–1994, 1995–1999 and 2000–2004 peri-ods, respectively. In addition, the variety of molecular types of fragments has increased over time, with Fab the only fragment type first studied in the clinic before 1996. Interestingly, of these Fab candidates, all except one were derived from murine mAbs and were ultimately discontinued. This sug-gests that projects that failed before 1996 did so for qualitatively different reasons than those that failed subsequently.

Approval successThe clinical phase–transition probabilities (probabilities of success, POS; Box 1) bench-marks are useful for strategic planning and investment of resources. Although POS values specific for antibody fragments have not been reported previously, our group5 has published approval success rates for full-length mAbs, and these were found to vary with such factors as sequence source (usually mouse or human)

neonatal Fc receptor–mediated recycling24,25. Rapid clearance can be desirable under some circumstances; for example, delivery of cyto-toxic radioisotopes. The risk, however, is that a short half-life may prevent sufficient accu-mulation of therapy at the targeted site26,27.

As fragments are relatively small, ungly-cosylated proteins, production in microbial expression systems may be easier and less costly than the conventional mammalian cell culture systems used to produce full-size mAbs and other biologics, although pub-licly available detailed economic analyses are limited10,28. That said, fragments are also more likely to form undesirable aggregates and can be less stable than full-size mAbs8, although the propensity for aggregation varies with fragment type29 and approaches to ameliorate the problem have been developed30,31.

Finally, fragment modes of action include blocking the action of biological molecules by binding either ligand or receptor, or engaging signaling pathways by cross-linking receptors. Unlike full-length mAbs, however, fragments (unless designed otherwise) fail to induce effector functions that require an Fc domain, such as antibody-dependent cell-mediated cytotoxicity or complement-dependent cyto-toxicity32.

Overview of fragment data setFragments make up a small percentage of mAb candidates evaluated in clinical studies. More than 450 mAbs have entered clinical study through commercial sponsors1, 54 of

nity. In addition, ‘third-generation’ antibodies can be produced in various ways10. Domains can be removed from antibodies, resulting in miniaturized antibodies10. Indeed, much as antigen-binding sequences can be selected for ‘humanization’, these sequences can be expressed as a miniaturized scFv. As an alter-native strategy, antigen-binding domains of immunoglobulins from other species have also been exploited, including that of the dromedary, whose antigen-binding domain is composed of a single heavy chain, the vari-able heavy chain (VHH)11.

Pros and cons of fragmentsAntibody fragments have both pros and cons as therapeutics compared with full-size mAb therapeutics. One advantage is that they are smaller and penetrate tissues and tumors more rapidly and deeply than mAbs12–14, which, because of such physiologic parameters as interstitial pressure15–17, tend to concentrate at the perivascular ‘binding site barrier’17,18. In addition, the small size of fragments has been suggested to permit binding to cryptic epitopes not accessible to full-sized mAbs13, including immuno-evasive pathogen glyco-proteins19 and enzyme active-site pockets20.

On the downside, fragments demonstrate short circulating half-lives in humans21, likely due to kidney clearance22,23, necessitating the need for derivitization with polyethylene gly-col (PEGylation)9, or other modifications. In contrast, full-length mAbs are not filtered by the kidney and are additionally protected by

Table 1 US approved human mAbs and antibody fragments

Trade name (generic name) Description Indication of first FDA approvalFDA designations

Date of first FDA approval Company

ReoPro (abciximab) Anti-GPIIb/IIIa, chimeric Fab, IgG1-

Prevention of blood clots in angioplasty

Priority 12/22/1994 Centocor/Eli Lilly

Lucentis (ranibizumab) Anti-VEGF-A, humanized Fab IgG1-

Wet age-related macular degeneration

Priority 6/30/2006 Genentech

Cimzia (certolizumab pegol) Anti-TNF- , PEGylated humanized Fab

Moderate to severe Crohn’s disease

Standard 4/22/2008 UCB

0

5

10

15

20

ApprovedDiscontinuedCurrently in studies

ApprovedPhase 3Phase 2Phase 1

8 9

2

9

18

4 4

Phase of developmentNum

ber

of a

ntib

ody

frag

men

ts

Figure 1 Phase of clinical development for 54 antibody fragment therapeutics.

F EATURE

Phase of clinical development for 54 antibody fragment therapeutics

Nelson AL, Reichert JM. Development trends for therapeutic antibody fragments. Nat Biotechnol. 2009 Apr.;27(4):331–337.


a trend comes from comparison of the per-centage of candidates that entered clinical study in the 1990s which were immunologi-cal agents (10%) to those that entered after 2000 (15%). A similar comparison of antine-oplastic candidates reveals that these agents made up 68% of all fragments entering the clinic during the 1990s, but 62% since 2000. Clinical studies of anti-infective candidates have also increased in number since 2000, up from 0% to 11%.

Fragment targetsTargets for fragments studied in the clinic were diverse—only ten antigens were tar-geted by more than one fragment, and the average number of fragments targeting the same antigen was only 2.9 (Table 2). Six tar-gets for multiple fragments (CD20, CD22, carcinoembryonic antigen (CEA), EpCAM, HER2/Neu and Lewis Y) are also common targets for full-length anticancer mAbs35; for example, EpCAM has been the target for at least 17 anticancer mAbs. Major non-oncologic targets include TNF- , which is targeted by four fragments in development and one marketed product (Cimzia, certoli-zumab pegol, a PEGylated anti-TNF ; UCB, Brussels, formerly Celltech) and platelet gly-coprotein (GP)-IIB/IIIA, which is targeted by two fragments in development and one mar-keted product (ReoPro, abciximab; Centocor, Horsham, PA, USA/Eli Lilly, Indianapolis).

As expected from the distribution of clini-cal indications, a substantial fraction of the 19 fragments currently in clinical study are directed against oncogenic pathways, especially when markers of hematologic malignancy are included. The percentage of fragments that bind tumor cell antigens was 48% and 35% for candidates that entered the clinic before 2000 and after 2000, respectively—a 13% decrease. In comparison, the percentage of fragments targeting antigens associated with immunol-ogy (immunity or inflammation) increased from 19% to 46% for the same periods. It is important to note that many of the antigenic targets associated with immunology are also potentially relevant to oncologic indications, and several fragments have been studied as treatments for both oncologic and immu-nological diseases (in such cases, we based our primary therapeutic category classifica-tion on the most advanced phase of study). Nevertheless, these data suggest that the over-all focus of drug developers is shifting from oncology toward immunological diseases.

Fragment immunoconjugatesFragments have fewer modes of action than full-size mAbs, unless they are specifically

ferent fragment types over time confirms the succession of Fabs, scFvs and third-genera-tion fragments. Fabs were the only fragment type that entered clinical study through the mid-1990s. Production of variable-region fragments capable of binding antigen, an antibody fragment type that includes scFv, was reported as early as 1988 (refs. 8,33,34), but scFv candidates entered clinical study only after 1995. During 1996 to 2000, scFvs accounted for 38% of the 16 fragments that entered the clinic. Since 2000, scFvs have made up 54% of 26 fragments introduced into clini-cal studies. Third-generation molecules began entering clinical studies during only the past five years.

Clinical indicationsMAbs have historically been studied most frequently as treatments for cancerous and immunological diseases5. This is also true for fragments, with 35 anticancer and 6 immu-nological candidates constituting 65% and 11% of the data set, respectively. Antibody fragments have also been developed for cardiovascular/hematologic disease (15%), infectious disease (6%) and ophthalmic con-ditions (4%). For discontinued programs, the percentage of antineoplastic fragments was slightly higher than the percentage of these agents in active clinical studies (71% versus 63%); in contrast, there was less clini-cal attrition for fragments developed for an immunologic indication (6% of discontinued fragments versus 16% of those in active clini-cal study). Immunological and antineoplastic fragments constitute nearly equal percentages of the preclinical pipeline (Fig. 3).

Our observations suggest a small shift in the focus of R&D programs from oncology to immunology. Further evidence for such

scFv fragment binding the Staphylococcus aureus ATP-binding cassette transporter (Novartis, Basel).

Fragment technologiesThree classes of mAb-derived fragments have entered clinical development—Fabs, scFvs and third-generation molecules. Analysis of the 54 agents in clinical development suggests that these technologies represent successive waves of technologies. This conclusion is supported by stratification of the data set by developmental fate (Fig. 2) and by time.

Thirty Fabs account for 56% of the frag-ment data set, with seven agents having pro-gressed beyond proof of concept, including three approved in the United States and one approved in China. Four Fabs account for 21% of open clinical development projects, and 22 account for 71% of all discontinued fragments. Fabs were the first type of anti-body fragments to enter clinical study. Of the discontinued Fabs, the majority (64%) were murine derived, with half first administered to humans before 1995.

Of the scFvs, 19 account for 35% of the entire fragment data set, 10 for 53% of those in ongoing studies, and 9 for 29% of discon-tinued fragments. Three scFvs cleared the proof-of-concept hurdle, but two were sub-sequently discontinued in phase 3.

The pipeline of fragments derived from third-generation technologies, including dromedary VHH–based single-chain, min-iaturized scFvs, and miniaturized antibody therapeutic molecules is the least mature. Only five third-generation molecules have entered clinical development under commercial spon-sorship, all of which are in early development. Trubion has sponsored the development of two anti-CD20 miniaturized mAbs for the treatment of rheumatologic conditions (TRU-015 and SBI-087) and an anti-CD37 minia-turized mAb for the treatment of leukemia (TRU-016). Arana has in-licensed Domantis’ miniaturized scFv technology to create ART-621, an anti-TNF fragment in development for immunologic disease. Finally, Ablynx has entered ALX-0081 into development for the treatment of coronary syndromes, the first dromedary-derived VHH molecule in clinical development.

Overall, Fabs were proportionally overrep-resented in the population of discontinued fragments, whereas scFv and third-generation molecules predominate among those in active clinical development (Fig. 2). An even greater emphasis on third-generation molecules was observed among preclinical agents. Similar trends were observed for preclinical agents. Analysis of trends in the introduction of dif-

0

10

20

30

40

50

60

70

80

90

100

3Gs

scFvs

Fabs

DiscontinuedClinicalPreclinical

21%

29%

50%

21%

53%

26%

71%

29%

Developmental category

Frag

men

t typ

e (%

)

Figure 2 Distribution of fragment type across developmental categories. 3G, third generation.

F EATURE


a trend comes from comparison of the per-centage of candidates that entered clinical study in the 1990s which were immunologi-cal agents (10%) to those that entered after 2000 (15%). A similar comparison of antine-oplastic candidates reveals that these agents made up 68% of all fragments entering the clinic during the 1990s, but 62% since 2000. Clinical studies of anti-infective candidates have also increased in number since 2000, up from 0% to 11%.

Fragment targetsTargets for fragments studied in the clinic were diverse—only ten antigens were tar-geted by more than one fragment, and the average number of fragments targeting the same antigen was only 2.9 (Table 2). Six tar-gets for multiple fragments (CD20, CD22, carcinoembryonic antigen (CEA), EpCAM, HER2/Neu and Lewis Y) are also common targets for full-length anticancer mAbs35; for example, EpCAM has been the target for at least 17 anticancer mAbs. Major non-oncologic targets include TNF- , which is targeted by four fragments in development and one marketed product (Cimzia, certoli-zumab pegol, a PEGylated anti-TNF ; UCB, Brussels, formerly Celltech) and platelet gly-coprotein (GP)-IIB/IIIA, which is targeted by two fragments in development and one mar-keted product (ReoPro, abciximab; Centocor, Horsham, PA, USA/Eli Lilly, Indianapolis).

As expected from the distribution of clini-cal indications, a substantial fraction of the 19 fragments currently in clinical study are directed against oncogenic pathways, especially when markers of hematologic malignancy are included. The percentage of fragments that bind tumor cell antigens was 48% and 35% for candidates that entered the clinic before 2000 and after 2000, respectively—a 13% decrease. In comparison, the percentage of fragments targeting antigens associated with immunol-ogy (immunity or inflammation) increased from 19% to 46% for the same periods. It is important to note that many of the antigenic targets associated with immunology are also potentially relevant to oncologic indications, and several fragments have been studied as treatments for both oncologic and immu-nological diseases (in such cases, we based our primary therapeutic category classifica-tion on the most advanced phase of study). Nevertheless, these data suggest that the over-all focus of drug developers is shifting from oncology toward immunological diseases.

Fragment immunoconjugatesFragments have fewer modes of action than full-size mAbs, unless they are specifically

ferent fragment types over time confirms the succession of Fabs, scFvs and third-genera-tion fragments. Fabs were the only fragment type that entered clinical study through the mid-1990s. Production of variable-region fragments capable of binding antigen, an antibody fragment type that includes scFv, was reported as early as 1988 (refs. 8,33,34), but scFv candidates entered clinical study only after 1995. During 1996 to 2000, scFvs accounted for 38% of the 16 fragments that entered the clinic. Since 2000, scFvs have made up 54% of 26 fragments introduced into clini-cal studies. Third-generation molecules began entering clinical studies during only the past five years.

Clinical indicationsMAbs have historically been studied most frequently as treatments for cancerous and immunological diseases5. This is also true for fragments, with 35 anticancer and 6 immu-nological candidates constituting 65% and 11% of the data set, respectively. Antibody fragments have also been developed for cardiovascular/hematologic disease (15%), infectious disease (6%) and ophthalmic con-ditions (4%). For discontinued programs, the percentage of antineoplastic fragments was slightly higher than the percentage of these agents in active clinical studies (71% versus 63%); in contrast, there was less clini-cal attrition for fragments developed for an immunologic indication (6% of discontinued fragments versus 16% of those in active clini-cal study). Immunological and antineoplastic fragments constitute nearly equal percentages of the preclinical pipeline (Fig. 3).

Our observations suggest a small shift in the focus of R&D programs from oncology to immunology. Further evidence for such

scFv fragment binding the Staphylococcus aureus ATP-binding cassette transporter (Novartis, Basel).

Fragment technologiesThree classes of mAb-derived fragments have entered clinical development—Fabs, scFvs and third-generation molecules. Analysis of the 54 agents in clinical development suggests that these technologies represent successive waves of technologies. This conclusion is supported by stratification of the data set by developmental fate (Fig. 2) and by time.

Thirty Fabs account for 56% of the frag-ment data set, with seven agents having pro-gressed beyond proof of concept, including three approved in the United States and one approved in China. Four Fabs account for 21% of open clinical development projects, and 22 account for 71% of all discontinued fragments. Fabs were the first type of anti-body fragments to enter clinical study. Of the discontinued Fabs, the majority (64%) were murine derived, with half first administered to humans before 1995.

Of the scFvs, 19 account for 35% of the entire fragment data set, 10 for 53% of those in ongoing studies, and 9 for 29% of discon-tinued fragments. Three scFvs cleared the proof-of-concept hurdle, but two were sub-sequently discontinued in phase 3.

The pipeline of fragments derived from third-generation technologies, including dromedary VHH–based single-chain, min-iaturized scFvs, and miniaturized antibody therapeutic molecules is the least mature. Only five third-generation molecules have entered clinical development under commercial spon-sorship, all of which are in early development. Trubion has sponsored the development of two anti-CD20 miniaturized mAbs for the treatment of rheumatologic conditions (TRU-015 and SBI-087) and an anti-CD37 minia-turized mAb for the treatment of leukemia (TRU-016). Arana has in-licensed Domantis’ miniaturized scFv technology to create ART-621, an anti-TNF fragment in development for immunologic disease. Finally, Ablynx has entered ALX-0081 into development for the treatment of coronary syndromes, the first dromedary-derived VHH molecule in clinical development.

Overall, Fabs were proportionally overrep-resented in the population of discontinued fragments, whereas scFv and third-generation molecules predominate among those in active clinical development (Fig. 2). An even greater emphasis on third-generation molecules was observed among preclinical agents. Similar trends were observed for preclinical agents. Analysis of trends in the introduction of dif-

0

10

20

30

40

50

60

70

80

90

100

3Gs

scFvs

Fabs


21%

29%

50%

21%

53%

26%

71%

29%


Frag

men

t typ

e (%

)Figure 2 Distribution of fragment type across developmental categories. 3G, third generation.

F EATURE

Distribution of fragment type across developmental categories


NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 4 APRIL 2009 335

The third approved fragment, Cimzia (cer-tolizumab pegol), is a PEGylated Fab targeting TNF- approved by FDA in April 2008 as a treatment for Crohn’s disease40. UCB’s Cimzia is the fourth anti-TNF- biologic on the market, competing with two full-sized mAbs (Remicade, (infliximab; Johnson & Johnson, New Brunswick, NJ, USA) and Humira (adali-mumab, Abbott)), a fusion protein (Enbrel, etanercept; Amgen, Thousand Oaks, CA, USA) and a fully human, full-sized mAb cur-rently under review by the FDA (golimumab, Johnson & Johnson).

Clinical candidatesNineteen fragments in the data set are actively in clinical studies. Three of the 19 active frag-ments included in the data set are bispecific molecules (16%), and one bivalent, monospe-cific fragment is in active development. The majority of active development programs are for scFv molecules (n = 10, 53%), of which 80% are for treatment of cancer. Four Fabs, three miniaturized antibodies, one miniatur-ized scFv, and one dromedary VHH–derived candidate make up the remaining active proj-ects, four of which are intended to treat cancer and three to treat immunologic disease. Of the 19 active projects, 12 are antineoplastic agents (63%), 3 are immunomodulatory (16%), 2 are anti-infective (11%) and the remaining 2 (5%) are in development for cardiovascular and ophthalmic indications. Most active fragment projects are in early-stage development—eight are in phase 1 and nine are in phase 2.

Two antineoplastic conjugate-fragment candidates have reached phase 3 studies. Viventia Biotech’s VB4-845 (an anti-EpCAM humanized scFv linked to Pseudomonas endo-

ReoPro (abciximab; Centocor) targets platelet glycoprotein (GP) IIb/IIIa and was approved by the FDA in December 1994 (refs. 36,37). The product prevents platelet aggrega-tion and thrombus formation and is used as an adjunct to coronary artery catheterization for ST-elevation myocardial infarction.

In June 2006, Lucentis (ranibizumab; Genentech), a Fab fragment targeting vascu-lar endothelial growth factor (VEGF)-A, was approved by FDA as a treatment for neovas-cular wet age-related macular degeneration38. Lucentis is similar, but not identical, to a Fab of Avastin (bevacizumab; Genentech), which is a full-size anti-VEGF-A humanized mAb approved as a treatment for colorectal cancer, and debate exists as to the relative merits and cost effectiveness of these agents39.

designed with additional features. One way to confer added functionality is to conjugate to fragments radioactive elements or other mol-ecules that either directly or indirectly affect targeted antigens. Of the 54 fragments in the data set, 24 candidates were constructed as immunoconjugates (44%), all of them as anti-cancer agents. The majority (69%) of the 35 anticancer fragments were thus immunocon-jugates, a higher fraction than that observed for the cohort of all anticancer mAbs, includ-ing fragments, studied in humans to date (44%)35.

In terms of current status, one radiolabeled Fab, Licartin (radioiodinated metuximab; Chengdu Hoist Hi-tech, Chengdu, China) has been approved for treatment of liver cancer in China (it has not been approved by the FDA). Eight of the 24 conjugated fragments (33%) are active projects currently in clinical study, including two agents in phase 3 trials. The 16 discontinued candidates were terminated in phase 1 or 2; 10 of these had entered clinical study before 2000. Six of 8 immunoconjugate fragments in clinical development are scFvs (75%), whereas only 6 of the 16 discontinued programs involved scFvs (38%), suggesting that the immunoconjugate fragment pipeline is cur-rently focused on development of scFvs.

In terms of format, of the 24 conjugated fragments entering the clinic, 12 were con-jugated to bacterial toxins: seven scFvs were fused to pseudomonas exotoxin; four Fabs were fused to staphylococcal enterotoxin A; and one Fab fragment was fused to bouga-nin, an rRNA N-glycosidase isolated from the Bougainvillea plant. Only 5 of these 12 can-didates currently remain in clinical study, but two have reached phase 3. Four frag-ment candidates were conjugated to other protein molecules, including interleukin (IL)-2, TNF- , epidermal growth factor (EGF) and the enzyme carboxypeptidase 2, which processes prodrugs into active cyto-toxic compounds. Two of these candidates are currently in clinical studies. The remaining seven fragments (of 24 immunoconjugates) incorporate radioisotopes, one of which has been approved in China (metuximab) and one (L19scFv; Philogen, Sienna, Italy), which remains in clinical study.

FDA-approved fragmentsIt is notable that, despite the focus on cancer treatments, none of the three FDA-approved fragments has been approved for cancer indi-cations. The products—ReoPro, Lucentis and Cimzia—are used as treatments for cardio-vascular, ophthalmic and immunological indications, respectively (Table 1). All three approved products are Fab molecules.

0

10

20

30

40

50

60

70

80

90

100

Ophthalmic

Cardiovascularhematologic

Anti-infective

Immunologic

Antineoplastic


46%

63%71%

41%16% 6%

3%11%

3%

19%8% 5%

5%3%

Ther

apeu

tic c

ateg

ory

(%)


Figure 3 Distribution of clinical indications across developmental categories. (Due to rounding, two bars do not add up to 100%.)

Table 2 Known targets of antibody fragment therapeutics that have entered clinical studya

Target (therapeutic category)Total number of

fragments

Fragments currently in

studies Fab onlyImmunoconjugate

fragments only

5T4 (cancer) 2 1 2 2

CD20 (cancer, immunological) 3 2 0 1

CD22 (cancer) 2 1 0 2

CEA (cancer) 3 0 2 3

EpCAM (cancer) 4 3 2 3

Fibronectin (cancer) 3 3 0 3

GPIIb/IIIa (cardiovascular) 3 0 3 0

HER2/neu (cancer) 3 1 1 1

Lewis Y (cancer) 2 0 0 2

TNF- (immunological) 4 2 2 0aIncludes 10 molecules that were targets for a minimum of two of the 54 antibody fragments in the data set. An additional 25 unique molecules were each targeted by one mAb fragment. The T-cell receptor (CD3) target and Fc R (CD64) target used by some bispecific candidates were not included, although the other disease-relevant targets for these bispecific fragments were included.

F EATURE

NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 4 APRIL 2009 335

The third approved fragment, Cimzia (cer-tolizumab pegol), is a PEGylated Fab targeting TNF- approved by FDA in April 2008 as a treatment for Crohn’s disease40. UCB’s Cimzia is the fourth anti-TNF- biologic on the market, competing with two full-sized mAbs (Remicade, (infliximab; Johnson & Johnson, New Brunswick, NJ, USA) and Humira (adali-mumab, Abbott)), a fusion protein (Enbrel, etanercept; Amgen, Thousand Oaks, CA, USA) and a fully human, full-sized mAb cur-rently under review by the FDA (golimumab, Johnson & Johnson).

Clinical candidatesNineteen fragments in the data set are actively in clinical studies. Three of the 19 active frag-ments included in the data set are bispecific molecules (16%), and one bivalent, monospe-cific fragment is in active development. The majority of active development programs are for scFv molecules (n = 10, 53%), of which 80% are for treatment of cancer. Four Fabs, three miniaturized antibodies, one miniatur-ized scFv, and one dromedary VHH–derived candidate make up the remaining active proj-ects, four of which are intended to treat cancer and three to treat immunologic disease. Of the 19 active projects, 12 are antineoplastic agents (63%), 3 are immunomodulatory (16%), 2 are anti-infective (11%) and the remaining 2 (5%) are in development for cardiovascular and ophthalmic indications. Most active fragment projects are in early-stage development—eight are in phase 1 and nine are in phase 2.

Two antineoplastic conjugate-fragment candidates have reached phase 3 studies. Viventia Biotech’s VB4-845 (an anti-EpCAM humanized scFv linked to Pseudomonas endo-

ReoPro (abciximab; Centocor) targets platelet glycoprotein (GP) IIb/IIIa and was approved by the FDA in December 1994 (refs. 36,37). The product prevents platelet aggrega-tion and thrombus formation and is used as an adjunct to coronary artery catheterization for ST-elevation myocardial infarction.

In June 2006, Lucentis (ranibizumab; Genentech), a Fab fragment targeting vascu-lar endothelial growth factor (VEGF)-A, was approved by FDA as a treatment for neovas-cular wet age-related macular degeneration38. Lucentis is similar, but not identical, to a Fab of Avastin (bevacizumab; Genentech), which is a full-size anti-VEGF-A humanized mAb approved as a treatment for colorectal cancer, and debate exists as to the relative merits and cost effectiveness of these agents39.

designed with additional features. One way to confer added functionality is to conjugate to fragments radioactive elements or other mol-ecules that either directly or indirectly affect targeted antigens. Of the 54 fragments in the data set, 24 candidates were constructed as immunoconjugates (44%), all of them as anti-cancer agents. The majority (69%) of the 35 anticancer fragments were thus immunocon-jugates, a higher fraction than that observed for the cohort of all anticancer mAbs, includ-ing fragments, studied in humans to date (44%)35.

In terms of current status, one radiolabeled Fab, Licartin (radioiodinated metuximab; Chengdu Hoist Hi-tech, Chengdu, China) has been approved for treatment of liver cancer in China (it has not been approved by the FDA). Eight of the 24 conjugated fragments (33%) are active projects currently in clinical study, including two agents in phase 3 trials. The 16 discontinued candidates were terminated in phase 1 or 2; 10 of these had entered clinical study before 2000. Six of 8 immunoconjugate fragments in clinical development are scFvs (75%), whereas only 6 of the 16 discontinued programs involved scFvs (38%), suggesting that the immunoconjugate fragment pipeline is cur-rently focused on development of scFvs.

In terms of format, of the 24 conjugated fragments entering the clinic, 12 were con-jugated to bacterial toxins: seven scFvs were fused to pseudomonas exotoxin; four Fabs were fused to staphylococcal enterotoxin A; and one Fab fragment was fused to bouga-nin, an rRNA N-glycosidase isolated from the Bougainvillea plant. Only 5 of these 12 can-didates currently remain in clinical study, but two have reached phase 3. Four frag-ment candidates were conjugated to other protein molecules, including interleukin (IL)-2, TNF- , epidermal growth factor (EGF) and the enzyme carboxypeptidase 2, which processes prodrugs into active cyto-toxic compounds. Two of these candidates are currently in clinical studies. The remaining seven fragments (of 24 immunoconjugates) incorporate radioisotopes, one of which has been approved in China (metuximab) and one (L19scFv; Philogen, Sienna, Italy), which remains in clinical study.

FDA-approved fragmentsIt is notable that, despite the focus on cancer treatments, none of the three FDA-approved fragments has been approved for cancer indi-cations. The products—ReoPro, Lucentis and Cimzia—are used as treatments for cardio-vascular, ophthalmic and immunological indications, respectively (Table 1). All three approved products are Fab molecules.

0

10

20

30

40

50

60

70

80

90

100

Ophthalmic

Cardiovascularhematologic

Anti-infective

Immunologic

Antineoplastic


46%

63%71%

41%16% 6%

3%11%

3%

19%8% 5%

5%3%

Ther

apeu

tic c

ateg

ory

(%)


Figure 3 Distribution of clinical indications across developmental categories. (Due to rounding, two bars do not add up to 100%.)

Table 2 Known targets of antibody fragment therapeutics that have entered clinical studya

Target (therapeutic category)Total number of

fragments

Fragments currently in

studies Fab onlyImmunoconjugate

fragments only

5T4 (cancer) 2 1 2 2

CD20 (cancer, immunological) 3 2 0 1

CD22 (cancer) 2 1 0 2

CEA (cancer) 3 0 2 3

EpCAM (cancer) 4 3 2 3

Fibronectin (cancer) 3 3 0 3

GPIIb/IIIa (cardiovascular) 3 0 3 0

HER2/neu (cancer) 3 1 1 1

Lewis Y (cancer) 2 0 0 2

TNF- (immunological) 4 2 2 0aIncludes 10 molecules that were targets for a minimum of two of the 54 antibody fragments in the data set. An additional 25 unique molecules were each targeted by one mAb fragment. The T-cell receptor (CD3) target and Fc R (CD64) target used by some bispecific candidates were not included, although the other disease-relevant targets for these bispecific fragments were included.

F EATURE

Distribution of clinical indications across developmental categories


Molecular classes of fragment therapeutics in preclinical research and clinical development - Fab

PC, preclinical; App, approved.


Molecular classes of fragment therapeutics in preclinical research and clinical development - scFv


are shifting from antineoplastic indications toward immunologic disease.

Fabs account for the majority of fragment candidates that have entered clinical develop-ment, and three have been approved by FDA. Even so, Fabs account for less than a quarter of fragments in ongoing clinical or preclini-cal development. Indeed, the percentage of the clinical pipeline accounted for by Fabs has fallen from 100% before 1995 to 37% in the 2000–2008 period. In contrast, scFvs cur-rently account for 53% of fragments in active development. The first scFvs entered clinical development in 1995, and have expanded in number to account for 44% of development projects during 2000–2008. Thus, scFv candi-dates appear to dominate mAb fragment clini-cal development, although it is important to note that scFvs are still a nascent therapeutic category—none is yet FDA approved and only one scFv is in phase 3 testing.

So-called third-generation antibody frag-ment technologies, dromedary VHH-based and miniaturized antibodies, are in the earli-est phases of clinical development, account-ing for only 9% of all fragments to enter the clinical pipeline but 50% of identified pre-clinical research. Indeed, these technologies are sufficiently new that little data on clinical candidates are available, making their future prospects difficult to predict.

The trend of a shift away from antineoplas-tic toward immunomodulatory therapies is demonstrated by the percentages of anti-neoplastic agents at the discontinued (71%), active (63%) and preclinical (45%) stages. This suggests that fewer antibody fragments for oncology are entering the clinic than before. In contrast, immunological candi-dates at the same stages show a reverse trend; more are in preclinical research (40%) than are active (16%), with only 6% discontinued in the clinic. The trend is also apparent when antigenic targets of fragments in the various categories of R&D are examined. The percent-

basis, the preclinical pipeline includes fewer anticancer candidates (45% preclinical ver-sus 63% clinical) and more candidates for immunological diseases (40% preclinical ver-sus 16% clinical) compared with the pipeline of fragments that has already entered clinical study (Fig. 3).

As noted, errors of omission among the preclinical fragment data set are likely, and data on the preclinical candidates identified were limited, as demonstrated by the fact that target antigens are definitively described for only 76% of fragments in the set. Even so, our data suggest that the preclinical pipeline may be enriched with multivalent and bispecific, third-generation fragments intended to treat oncologic and immunologic diseases.

Development trendsThe pharmaceutical industry has become increasingly interested in therapeutic mAbs in general, and antibody fragments in particular. In the past, fragment technologies were over-shadowed by the currently more successful full-length mAb technologies, which together have generated 19 FDA approved products to date and more than 200 candidates in clinical development1. Even so, fragments offer sev-eral advantages over full-length mAbs and also embody the growing diversity of recom-binant biologic therapeutics that draw on protein domains from any source. Advances in technology have driven the current focus on study of antibody fragments, although this is a nascent category, and the safety and efficacy of most candidates have yet to be determined.

By comparing the prevalence of each frag-ment type in active clinical development to its prevalence among discontinued agents and those in preclinical research, we have observed two notable trends: successive waves of different antibody fragment technologies is reflected in the numbers of Fabs, scFvs, and third-generation candidates entering clini-cal trials at different times; and R&D efforts

toxin) was discontinued in phase 3 as a treat-ment for intratumor treatment of head and neck cancer under the trade name Proxinium, but remains in phase 2 for intravesicular treat-ment of bladder cancer under the trade name Vicinium. Active Biotech is also developing a murine Fab, Anyara (naptumomab estafena-tox), for the treatment of renal cell carcinoma that is currently in phase 3 trials.

Preclinical pipelineThe data set included 38 molecules in com-mercial preclinical research, which suggests that fragments are considered a promising area of therapeutic development. The number is an underestimate because many preclini-cal projects are described poorly, if at all, in publications. It should also be noted that data for fragments in preclinical development by academia and government institutions were excluded because the commercial potential for the molecules could not be assessed, although some of these candidates might be licensed to commercial firms in the future. Despite these limitations, the data set captures some of the key dimensions in fragment research at this time, and permits extrapolation about com-ing trends in technology development and clinical applications.

The 38 fragments in the preclinical pipe-line included 8 Fabs (21%), 11 scFvs (29%), 11 miniaturized scFvs (29%), 5 dromedary-based single-chain molecules (13%) and 3 miniaturized antibodies (8%) (Fig. 4). Almost half of these molecules are multivalent (42%), a notably greater fraction than observed in the clinical pipeline (19%), with bivalent, triva-lent and tetravalent molecules in preclini-cal studies. Fourteen are bispecific (36%), a higher rate than observed in the clinical pipeline (17%). Of the preclinical fragments in the data set, eight are conjugated molecules (21%), including three toxin conjugates, one cytokine conjugate, two drug conjugates and a radioisotope conjugate. On a percentage

Fab (n = 38) 3G (n = 24)scFv (n = 30)

Fab2

(~55 kDa) VhH domain(~15 kDa)

scFv(~28 kDa)

tascFv(~55 kDa)

Minibodybivalent

(~75 kDa)Bis-scFv(~55 kDa)

Fab2

(~110 kDa)

Fabbispecific

(~110 kDa)

Developmental category Developmental categoryDevelopmental categoryNum

ber

of th

erap

eutic

s

Num

ber

of th

erap

eutic

s

Num

ber

of th

erap

eutic

s

0

5

10

15

20

App321PC0

5

10

15

20

Discontinued Clinical

App321PC 0

5

10

15

20

App321PC

811

5 4110 0 0 0

4 3 2

19

2 30 000 0 0 00

5

15

22 14

Figure 4 Molecular classes of fragment therapeutics in preclinical research and clinical development. PC, preclinical; App, approved; 3G, third generation; tascFv, tandem single chain variable fragment; Bis, bispecific.

F EATURE

PC, preclinical; App, approved; tascFv, tandem single chain variable fragment; Bis, bispecific


Molecular classes of fragment therapeutics in preclinical research and clinical development - 3rd generation (3G)


are shifting from antineoplastic indications toward immunologic disease.

Fabs account for the majority of fragment candidates that have entered clinical develop-ment, and three have been approved by FDA. Even so, Fabs account for less than a quarter of fragments in ongoing clinical or preclini-cal development. Indeed, the percentage of the clinical pipeline accounted for by Fabs has fallen from 100% before 1995 to 37% in the 2000–2008 period. In contrast, scFvs cur-rently account for 53% of fragments in active development. The first scFvs entered clinical development in 1995, and have expanded in number to account for 44% of development projects during 2000–2008. Thus, scFv candi-dates appear to dominate mAb fragment clini-cal development, although it is important to note that scFvs are still a nascent therapeutic category—none is yet FDA approved and only one scFv is in phase 3 testing.

So-called third-generation antibody frag-ment technologies, dromedary VHH-based and miniaturized antibodies, are in the earli-est phases of clinical development, account-ing for only 9% of all fragments to enter the clinical pipeline but 50% of identified pre-clinical research. Indeed, these technologies are sufficiently new that little data on clinical candidates are available, making their future prospects difficult to predict.

The trend of a shift away from antineoplas-tic toward immunomodulatory therapies is demonstrated by the percentages of anti-neoplastic agents at the discontinued (71%), active (63%) and preclinical (45%) stages. This suggests that fewer antibody fragments for oncology are entering the clinic than before. In contrast, immunological candi-dates at the same stages show a reverse trend; more are in preclinical research (40%) than are active (16%), with only 6% discontinued in the clinic. The trend is also apparent when antigenic targets of fragments in the various categories of R&D are examined. The percent-

basis, the preclinical pipeline includes fewer anticancer candidates (45% preclinical ver-sus 63% clinical) and more candidates for immunological diseases (40% preclinical ver-sus 16% clinical) compared with the pipeline of fragments that has already entered clinical study (Fig. 3).

As noted, errors of omission among the preclinical fragment data set are likely, and data on the preclinical candidates identified were limited, as demonstrated by the fact that target antigens are definitively described for only 76% of fragments in the set. Even so, our data suggest that the preclinical pipeline may be enriched with multivalent and bispecific, third-generation fragments intended to treat oncologic and immunologic diseases.

Development trendsThe pharmaceutical industry has become increasingly interested in therapeutic mAbs in general, and antibody fragments in particular. In the past, fragment technologies were over-shadowed by the currently more successful full-length mAb technologies, which together have generated 19 FDA approved products to date and more than 200 candidates in clinical development1. Even so, fragments offer sev-eral advantages over full-length mAbs and also embody the growing diversity of recom-binant biologic therapeutics that draw on protein domains from any source. Advances in technology have driven the current focus on study of antibody fragments, although this is a nascent category, and the safety and efficacy of most candidates have yet to be determined.

By comparing the prevalence of each frag-ment type in active clinical development to its prevalence among discontinued agents and those in preclinical research, we have observed two notable trends: successive waves of different antibody fragment technologies is reflected in the numbers of Fabs, scFvs, and third-generation candidates entering clini-cal trials at different times; and R&D efforts

toxin) was discontinued in phase 3 as a treat-ment for intratumor treatment of head and neck cancer under the trade name Proxinium, but remains in phase 2 for intravesicular treat-ment of bladder cancer under the trade name Vicinium. Active Biotech is also developing a murine Fab, Anyara (naptumomab estafena-tox), for the treatment of renal cell carcinoma that is currently in phase 3 trials.

Preclinical pipelineThe data set included 38 molecules in com-mercial preclinical research, which suggests that fragments are considered a promising area of therapeutic development. The number is an underestimate because many preclini-cal projects are described poorly, if at all, in publications. It should also be noted that data for fragments in preclinical development by academia and government institutions were excluded because the commercial potential for the molecules could not be assessed, although some of these candidates might be licensed to commercial firms in the future. Despite these limitations, the data set captures some of the key dimensions in fragment research at this time, and permits extrapolation about com-ing trends in technology development and clinical applications.

The 38 fragments in the preclinical pipe-line included 8 Fabs (21%), 11 scFvs (29%), 11 miniaturized scFvs (29%), 5 dromedary-based single-chain molecules (13%) and 3 miniaturized antibodies (8%) (Fig. 4). Almost half of these molecules are multivalent (42%), a notably greater fraction than observed in the clinical pipeline (19%), with bivalent, triva-lent and tetravalent molecules in preclini-cal studies. Fourteen are bispecific (36%), a higher rate than observed in the clinical pipeline (17%). Of the preclinical fragments in the data set, eight are conjugated molecules (21%), including three toxin conjugates, one cytokine conjugate, two drug conjugates and a radioisotope conjugate. On a percentage

Fab (n = 38) 3G (n = 24)scFv (n = 30)

Fab2

(~55 kDa) VhH domain(~15 kDa)

scFv(~28 kDa)

tascFv(~55 kDa)

Minibodybivalent

(~75 kDa)Bis-scFv(~55 kDa)

Fab2

(~110 kDa)

Fabbispecific

(~110 kDa)

Developmental category Developmental categoryDevelopmental categoryNum

ber

of th

erap

eutic

s

Num

ber

of th

erap

eutic

s

Num

ber

of th

erap

eutic

s

0

5

10

15

20

App321PC0

5

10

15

20

Discontinued Clinical

App321PC 0

5

10

15

20

App321PC

811

5 4110 0 0 0

4 3 2

19

2 30 000 0 0 00

5

15

22 14

Figure 4 Molecular classes of fragment therapeutics in preclinical research and clinical development. PC, preclinical; App, approved; 3G, third generation; tascFv, tandem single chain variable fragment; Bis, bispecific.

F EATURE

PC, preclinical; App, approved; 3G, third generation.


The approved products accounted for ~$20 billion of healthcare spending in 2006, and the market is growing at ~14% per year. !Eight mAbs had sales >$1 billion in 2007, with Rituxan (rituximab; Genentech, S. San Francisco, CA, USA) being the bestseller, with sales of $5.2 billion4.


€ 650 per year of Denosumab (post-menopausal osteoporosis)

€ 550.000 per year of Eculizumab (paroxysmal nocturnal hemoglobinuria, PNH)

€ 7.500 to 13.000 per year Tocilizumab and Rituximab (Rheumatoid Arthritis)

€ 130.000 for Brentuximab vedotin, 1 year of treatment for a 70 kg patient (anti-neoplastic mAbs)

Cost of therapeutic mAbs (Italy)

Salemi S1, Markovic M, Martini G, D'Amelio R. The Expanding Role of Therapeutic Antibodies. Int Rev Immunol. 2014 Jan 28. [Epub ahead of print]

contrast relatively to the intermediate clearing (I253A andH310A) and the fast clearing (H310A/H435Q) scFv-Fc frag-ments. The blood clearance of the scFv-Fc H310A/H435Q,also referred to as scFv-Fc double mutant (scFv-Fc DM), is

shown in Fig. 2A. This fragment exhibits longer persistance inthe circulation than the minibody which results in higher tumoruptake as seen in Fig. 2B. The PET images of 124I- and64Cu-DOTA-scFv-Fc also show high, specific tumor up-takes in xenografts (Fig. 5A). The pharmacokinetics ofradioiodinated versus radiometal-labeled I253A, H310Aand H310A/H435Q scFv-Fc variants were further evalu-ated in xenografted mice to predict their therapeutic po-tential.19 Tumor uptakes were inversely related to bloodclearance, and hepatic radiometal activity correlated withthe blood clearance rate of the fragment; that is fasterclearance resulted in higher activity. On the basis of thebiodistribution data with 125I and 111In, it was predictedthat the fast clearing scFv-Fc DM would be able to deliver!7000 cGy to the tumor, with favorable tumor to liverand kidney ratios when radiolabeled with 131I, whereas asfor 90Y therapy, a slow clearing antibody would be theprotein of choice as the liver or kidney activities would belower.

The scFv-Fc DM fragment has also been evaluated in othertumor models. An anti-HER2 scFv-Fc DM fragment was eval-uated by PET in xenografted mice after conjugation withDOTA and radiolabeling with 64Cu.95 The scFv-Fc DM frag-ment exhibited over 2-fold improved tumor targeting andreduced kidney activity relatively to the anti-HER2 minibodyfragment, which was probably due to its slightly longer resi-dence time in the blood. When this fragment was labeledwith 124I however, very low activity was observed in the tu-mor. In another recent study, an anti-CD20 scFv-Fc DMfragment with IgG4 Fc was evaluated in tumor bearing miceafter radioiodination with 124I (Fig. 5B).99 Although rapidtumor localization was observed, the activities in the tumorand the blood were 2-fold lower than of the minibody. Suchdifference was not observed with the anti-CEA (Fig. 2A) andanti-HER2 fragments. Thus, it appears to be related to thechange of IgG Fc isotype. Intact human IgG4 has been re-ported to clear faster than human IgG1 in both humans andmice.107,108 Consistent with this, a lower background activityof the radioiodinated anti-CD20 scFv-Fc DM fragment thanthat of the anti-CEA scFv-Fc DM at 18-20 hours p.i. wasobserved. Thus, the introduced mutations in IgG4 Fc mayaccelerate the scFv-Fc fragment to clear faster than the mini-body. Figure 5B also shows the images obtained with 64Cu-DOTA scFv-Fc DM fragment.

A rapid method to generate scFv-Fc fragments targetingendothelial cell from binders isolated from phage displaylibraries and in vivo evaluation by SPECT have been de-scribed.109 Using this approach, 3 antibodies were identifiedthat, when converted to scFv-Fc using mouse IgG1 Fc, tar-geted the lungs in rats with uptakes of 30%-40% ID/g at 1and 2 hours p.i. Because the blood activities were less than2% ID/g, high-contrast images were achieved early.

Impact of AntibodyModifications on ImagingFrom the numerous studies done with the different engi-neered antibody fragments, there are several factors, such as

21 h

64Cu 124I

18 h64Cu 124I

20 h18 h

A

B

Figure 5 Small animal PET imaging of xenograft-bearing mice in-jected with 64Cu and/or 124I-labeled scFv-Fc double mutant (DM)fragments targeting CEA (A) and CD20 (B).99,105 The backgroundactivity of 124I-anti-CD20 scFv-Fc with IgG4 Fc is less (1.9% ID/g at21 hours p.i.)99 than that seen with 124I-anti-CEA scFv-Fc with IgG1Fc region (6.9% ID/g at 12 hours and 2.8% ID/g at 24 hours p.i.),19

suggesting a more rapid clearance rate of this anti-CD20 fragment.This is consistent with published observation of the different bloodclearances of human IgG1 and IgG4 isotypes. Again, 64Cu-labeledfragments produce less contrast than 124I-labeled fragments, andhepatic clearance is evident.

176 T. Olafsen and A.M. Wu

Antibody constructs for imagingOlafsen T, Wu AM. Antibody vectors for imaging. Semin Nucl Med. 2010 May;40(3):167-181.

• Molecular imaging offers the ability to visualize, characterize, and measure processes on molecular and cellular levels noninvasively in living systems

• The key players for obtaining this information are the molecular imaging agent (probe or tracer) and the target, which can be intracellular or cell surface proteins

• Radiolabeled probes for SPECT (Single-Photon Emission Computed Tomography) or PET (Positron Emission Tomography), offer visualization of physiological and biochemical changes

• In contrast, conventional imaging modalities (radiography, ultrasonography, CT, and magnetic resonance imaging (MRI)) offer visualization of nonspecific changes related to morphology

The discipline of molecular imaging is one of the most rapidly growing areas of science

Olafsen T, Wu AM. Antibody vectors for imaging. Semin Nucl Med. 2010 May;40(3):167–181.

body fragments, such as diabodies (dimers of scFv, 50 kDa),minibodies (dimers of scFv-CH3, 80 kDa), and scFv-Fc (crys-tallizing fragment) dimers (105 kDa), have been generated(Fig. 1).16 These fragments have exhibited better tumor pen-etration, faster clearance kinetics, and excellent tumor toblood ratios, which are desirable properties for an imagingprobe. The blood clearance and tumor uptake curves of ra-dioiodinated anti-carcinoembryonic antigen (CEA) scFv,diabody, minibody, and a rapid clearing scFv-Fc fragmentare shown in Fig. 2. The rapid elimination has been attrib-uted to their size and lack of interaction with the neonatal Fcreceptor (FcRn).

After the optimal antibody-based vector has been gener-ated, they can be tagged with radionuclides (PET, SPECT),magnetic nanoparticles (MRI), or fluorescence/biolumines-cence (optical) probes that enable a variety of imaging mo-dalities. Although PET is more sensitive than SPECT, thelatter offer advantages, such as broad availability and lowercosts of radioisotopes and !-scanning instruments. In addi-tion, since SPECT radionuclides emit photons with differentenergies, several biomarkers can potentially be distinguishedat the same time, as opposed to PET radionuclides, which allemit the same energy photons (511 keV). MRI is a powerfulimaging modality with regard to high spatial resolution("100 #m) and tomographic capabilities, but the low signalsensitivity has been a major limitation for molecular imaging.However, the recent development and the increased sensitiv-ity of magnetism-engineered iron oxide nanoprobes obtainedwhen conjugated to Herceptin may make antibody fragmentsattractive MRI imaging agents.20 The extremely high sensitiv-ity of optical imaging approaches (10!15-10!17 mol/L for bi-oluminescence, 10!9-10!12 mol/L for fluorescence21) for real-time imaging of small animal models provide low-costalternatives to nuclear medicine imaging approaches. Themajor disadvantage of optical imaging using visible light ispoor tissue penetration ("1 cm) due to scatter and absorp-tion of light. Bioluminescence has some advantages over flu-orescence, since signal is produced in the presence of sub-strate (no excitation light source is needed), and there is nobackground autofluorescence. However, for optical imaging,

tomographic imaging remains a challenge. This review fo-cuses on the use of engineered antibody fragments as probesor vectors for noninvasive imaging of tumors, with an em-phasis on nuclear medicine applications.

Recombinant AntibodyFragments for ImagingMonovalent Single-Chain Fv FragmentsThe scFv fragment consists of variable light (VL) and heavy(VH) domains joined by a flexible peptide linker that can be12-25 residues long (Fig. 1). The resulting protein beingone-sixth of the size of an intact antibody (150 kDa), retainsthe specificity and affinity of the parental antibody. Althoughthe V domains can be in either orientation,that is, VL-linker-VH

or VH-linker-VL, the orientation can affect expression,22,23 sta-bility, and antigen binding activity.24 Early biodistributionand imaging studies in planar gamma cameras revealed thatalthough scFv molecules cleared very rapidly from blood,they were able to localize efficiently to their target.17,25-29

However, issues with direct iodination, such as loss of immu-noreactivity and deiodination, reduce the signal in the targettissue. Therefore, scFv molecules have been radiolabeledwith radiometals and other radionuclides. Several groupshave used 99mTc for 2 reasons: (1) its short physical half-life(6 hours) is compatible with the short biological half-life ofscFv, and (2) carrier-free 99mTc is generator produced andreadily available in hospital nuclear medicine departments.Genetic engineering approaches for site-specific 99mTc-label-ing have been created by C-terminal addition of a cysteinylpeptide (G)4C30 or a cys-his6 peptide31 for chelation chemis-try, metallothionein-myc peptide for direct radiolabeling,29

and His6 tag for using Tc(I)-carbonyl complex.32 Radiometallabeling of scFv with 177Lu and 111In after conjugation toDOTA or DTPA, respectively33,34 resulted in elevated kidneyretention and only a few percentage ID/g in the tumor inxenograft-bearing mice. When a disulfide stabilized anti-TacscFv (dsFv) radiolabeled with 18F was evaluated in ATAC4tumor bearing mice, maximum retention in the tumor was

Fast PK Slow PK

Intact an!body150 kDa

scFv-Fc105 kDa

Minibody80 kDa

Diabody50 kDa

scFv25 kDa

Fv12 kDa

VH

CH2

CH3

CH1

CL

VL

Fc

F(ab’)2

Figure 1 Schematic presentation of an intact antibody and engineered antibody fragments derived from it, including asingle variable domain fragment (Fv), single chain Fv (scFv), diabody, minibody, and scFv-Fc. Molecular weights areindicated below each fragment. (VL, variable light [light green]; VH, variable heavy [dark green]; CL, constant light[purple]; CH, constant heavy [blue]).

Antibody vectors for imaging 169


PharmacoKinetic


After the optimal antibody-based vector has been generated, they can be tagged with radionuclides (PET, SPECT), magnetic nanoparticles (MRI), or fluorescence/bioluminescence (optical) probes that enable a variety of imaging modalities


• The major limitations of scFv molecules are low functional affinity and short in vivo half-life due to their valency and small size, respectively

• Consequently, rapid dissociation from the target antigen due to monovalent binding will result in modest retention time in the target and potentially poor image quality

scFv limitations in molecular imaging

4.2% ID/g at 45 minutes p.i., and maximum retention in thekidneys was 303% ID/g at 15 minutes p.i.35 Thus, kidneys arethe dose-limiting organ of this fragment. Fluorescence imag-ing using fluorescent-conjugated scFv molecules for nonra-dioactive immune detection of tumors has also been investi-gated.36,37 Using cyanine fluorochromes Cy7 (infrared) andCy5.5.18 (far infrared), rapid secretion through the renal routewas observed early, whereas strong accumulation of fluores-cence in tumor was evident at 24 hours p.i.

The major limitations of scFv molecules are low functionalaffinity and short in vivo half-life (Fig. 2) due to their valency

and small size, respectively. Consequently, rapid dissociationfrom the target antigen due to monovalent binding will resultin modest retention time in the target and potentially poorimage quality.38 Still they remain attractive candidates, asthey can easily and cost effectively be expressed in bacteria.To extend the in vivo half-life, scFv molecules have beenconjugated to polyethylene glycol (PEG) polymers and hu-man serum albumin (HSA). After site-specific PEGylation of5, 20, and 40 kDa maleimide–PEG polymers to a scFv, pro-longed biological half-lives in vivo correlating to the molec-ular mass of the polymer was observed.39 HSA has a serumhalf-life of 19 days, and Yazaki et al40 fused HSA to anti-CEAT84.66 scFv molecule and evaluated the fusion molecule byboth SPECT and PET using different radionuclides. SPECTimages with 125I and 111In-DOTA-labeled scFv-HSA demon-strated rapid clearance and excellent tumor uptake. PET im-aging was carried out with 1 tumor bearing mouse injectedwith 64Cu-DOTA-scFv-HSA. In this animal, tumor localiza-tion was evident at 4 hours, which reached highest intensityat 24 hours p.i. The anti-CEA scFv-HSA fusion proteinprimary clearance route was through the liver (molecularweight ! 90 kDa). It exhibited significant slower bloodclearance and a higher tumor uptake than the minibody.18

These results are encouraging and be used as an alternativeway to overcome the very rapid clearance of scFv fragments.

Di- and MultivalentSingle-Chain Fv FragmentsThe monovalency of scFv is a significant limitation to tumorretention, and numerous approaches to genetically engineermonovalent scFv into multivalent fragments with greateravidity have been pursued. Joining 2 anti-tumor-associatedglycoprotein (TAG)-72 CC49 scFv molecules in tandem toproduce divalent sc(Fv)2 have resulted in spontaneousdimerization to a tetravalent (sc[Fv]2)2 with affinity and tu-mor retention similar to the intact parental antibody, butwith a more rapid blood clearance.41 In vivo studies with99mTc-CC49 sc(Fv)2 and (sc[Fv]2)2 demonstrated longerblood clearance and higher tumor retention with the tetrava-lent (sc[Fv]2)2.42 Chemical cross-linking of anti-TAG B72.3scFv molecules using bis-maleimide cross-linkers have beenemployed to produce dimers or trimers.43 Both exhibitedrapid clearance and similar tumor uptakes (approximately10% ID/g at 24 hours p.i.) in vivo, but high kidney retentionwas observed only with the radiometal labeled dimer. Inanother study, SPECT imaging of tumor bearing mice in-jected with a covalent linked 125I-sc(Fv)2 (divalent) againstplacental alkaline phosphatase, showed improved tumor up-take and longer retention time in the tumor compared with themonomeric scFv.44 Similarly, improved tumor uptake over the741F8 scFv monomer was achieved when it was made into adimer, bridged through a C-terminal gly4cys2 peptide.45

Use of multimerization domains or sequences to producehigh avidity scFv molecules has also been explored. Strepta-vidin-scFv fusion for production of tetravalent and bifunc-tional scFv molecules46,47 has largely been exploited for invivo pretargeting approaches for delivery of therapeutics

Figure 2 Blood clearance (A) and tumor uptake (B) curves of radio-iodine-labeled anti-carcinoembryonic antigen (CEA) T84.66 anti-body fragments in mice carrying subcutaneous LS174T human co-lon carcinoma xenografts.17-19 Standard errors are shown. Uptake isexpressed as percentage of injected dose per gram (% ID/g). ThescFv-Fc fragment, modified to clear quickly by the introduction of 2mutations (H310A/H435Q) in the Fc region that interfere the neo-natal Fc receptor interaction, is referred to as double mutant (DM).The longer persistence of radiolabeled fragments in the circulationleads to higher and more persistent tumor uptakes, but longer timesare required to obtain high contrast imaging.


4.2% ID/g at 45 minutes p.i., and maximum retention in thekidneys was 303% ID/g at 15 minutes p.i.35 Thus, kidneys arethe dose-limiting organ of this fragment. Fluorescence imag-ing using fluorescent-conjugated scFv molecules for nonra-dioactive immune detection of tumors has also been investi-gated.36,37 Using cyanine fluorochromes Cy7 (infrared) andCy5.5.18 (far infrared), rapid secretion through the renal routewas observed early, whereas strong accumulation of fluores-cence in tumor was evident at 24 hours p.i.

The major limitations of scFv molecules are low functionalaffinity and short in vivo half-life (Fig. 2) due to their valency

and small size, respectively. Consequently, rapid dissociationfrom the target antigen due to monovalent binding will resultin modest retention time in the target and potentially poorimage quality.38 Still they remain attractive candidates, asthey can easily and cost effectively be expressed in bacteria.To extend the in vivo half-life, scFv molecules have beenconjugated to polyethylene glycol (PEG) polymers and hu-man serum albumin (HSA). After site-specific PEGylation of5, 20, and 40 kDa maleimide–PEG polymers to a scFv, pro-longed biological half-lives in vivo correlating to the molec-ular mass of the polymer was observed.39 HSA has a serumhalf-life of 19 days, and Yazaki et al40 fused HSA to anti-CEAT84.66 scFv molecule and evaluated the fusion molecule byboth SPECT and PET using different radionuclides. SPECTimages with 125I and 111In-DOTA-labeled scFv-HSA demon-strated rapid clearance and excellent tumor uptake. PET im-aging was carried out with 1 tumor bearing mouse injectedwith 64Cu-DOTA-scFv-HSA. In this animal, tumor localiza-tion was evident at 4 hours, which reached highest intensityat 24 hours p.i. The anti-CEA scFv-HSA fusion proteinprimary clearance route was through the liver (molecularweight ! 90 kDa). It exhibited significant slower bloodclearance and a higher tumor uptake than the minibody.18

These results are encouraging and be used as an alternativeway to overcome the very rapid clearance of scFv fragments.

Di- and MultivalentSingle-Chain Fv FragmentsThe monovalency of scFv is a significant limitation to tumorretention, and numerous approaches to genetically engineermonovalent scFv into multivalent fragments with greateravidity have been pursued. Joining 2 anti-tumor-associatedglycoprotein (TAG)-72 CC49 scFv molecules in tandem toproduce divalent sc(Fv)2 have resulted in spontaneousdimerization to a tetravalent (sc[Fv]2)2 with affinity and tu-mor retention similar to the intact parental antibody, butwith a more rapid blood clearance.41 In vivo studies with99mTc-CC49 sc(Fv)2 and (sc[Fv]2)2 demonstrated longerblood clearance and higher tumor retention with the tetrava-lent (sc[Fv]2)2.42 Chemical cross-linking of anti-TAG B72.3scFv molecules using bis-maleimide cross-linkers have beenemployed to produce dimers or trimers.43 Both exhibitedrapid clearance and similar tumor uptakes (approximately10% ID/g at 24 hours p.i.) in vivo, but high kidney retentionwas observed only with the radiometal labeled dimer. Inanother study, SPECT imaging of tumor bearing mice in-jected with a covalent linked 125I-sc(Fv)2 (divalent) againstplacental alkaline phosphatase, showed improved tumor up-take and longer retention time in the tumor compared with themonomeric scFv.44 Similarly, improved tumor uptake over the741F8 scFv monomer was achieved when it was made into adimer, bridged through a C-terminal gly4cys2 peptide.45

Use of multimerization domains or sequences to producehigh avidity scFv molecules has also been explored. Strepta-vidin-scFv fusion for production of tetravalent and bifunc-tional scFv molecules46,47 has largely been exploited for invivo pretargeting approaches for delivery of therapeutics

Figure 2 Blood clearance (A) and tumor uptake (B) curves of radio-iodine-labeled anti-carcinoembryonic antigen (CEA) T84.66 anti-body fragments in mice carrying subcutaneous LS174T human co-lon carcinoma xenografts.17-19 Standard errors are shown. Uptake isexpressed as percentage of injected dose per gram (% ID/g). ThescFv-Fc fragment, modified to clear quickly by the introduction of 2mutations (H310A/H435Q) in the Fc region that interfere the neo-natal Fc receptor interaction, is referred to as double mutant (DM).The longer persistence of radiolabeled fragments in the circulationleads to higher and more persistent tumor uptakes, but longer timesare required to obtain high contrast imaging.


%ID/g = percentage of injected dose per gram


with 64Cu-DOTA diabody (Fig. 3A). The anti-CEA diabodywas also radiolabeled with 18F using N-succinimidyl-4-18F-fluorobenzoate, and evaluated by PET imaging in xenograftedmice.76 Clear delineation of the positive tumor was obtained asearly as 1 hour after injection of the 18F-FB-T84.66 diabody,with clearance predominantly through the kidneys. This studysuggests that the diabody is compatible with a short-lived radio-nuclide for same day imaging.

Enzymes that can catalyze a light-producing reaction (aprocess called bioluminescence) are known as luciferases.The light is produced after injection of a substrate for theenzyme. Luciferases are mainly used as reporter genes, withfirefly luciferase being the most commonly used for imag-

ing.79 Other luciferases, such as Renilla luciferase (RLuc) andGaussia luciferase (GLuc) from the sea pansy Renilla renifor-mis marine and copepod Gaussia princeps, respectively, havebeen fused to the anti-CEA diabody and evaluated by opticalbioluminescence imaging (Fig. 3B).77,78 Signal in the tumorwas observed from 2 to 24 hours with the diabody-RLuc8variant with the highest intensity at 4-8 hours p.i.78 Thetumors in mice injected with diabody-GLuc!15 also exhib-ited highest intensity at 4 hours p.i.77 However, although thediabody-GLuc!15 was brighter than diabody-RLuc8 invitro, the overall in vivo performance of diabody-GLuc!15was poorer than that of the diabody-RLuc8 with regard tostability, tumor uptake, and sustained targeting out to 24

64Cu 18F

4 h 4 h

124I

20 h

4 h 4 h4 h

Db-GL 15Db-RLuc8RLuc8

A

B

Figure 3 Small animal positron emission tomography (PET) (A) and optical imaging (B) of anti-CEA T84.66 diabody inLS174T xenograft-bearing mice.75-78 (A) Coronal PET images of mice injected with diabody after radiolabeling withdifferent positron emitting radionuclides. Specific localization to the positive tumor (arrow) and no localization to thenegative C6 rat glioma tumors (arrowhead) are seen in all the images. Renal clearance is evident at 4 hours p.i. with64Cu- and 18F-labeled diabodies, which is expected for a fragment of "60 kDa in size. High-contrast imaging is achieved with124I-labeled diabody at 20 hour p.i. due to almost complete elimination of any background activity, and a PET/computedtomography overlaid image is shown for anatomical reference. (B) Bioluminescence imaging of mice injected with diabody-Renilla (Db-RLuc8) and -Gaussia (Db-GL!15) luciferase fusion proteins at 4 hours p.i. The control mouse injected withRluc8 only is shown. Specific targeting to the positive tumor is seen with both fusion proteins.





64Cu 18F

4 h 4 h

124I

20 h

4 h 4 h4 h


A

B






64Cu 18F

4 h 4 h

124I

20 h

4 h 4 h4 h


A

B



Diabody


agent with rapid high-level accumulation of activity in xeno-grafts and rapid disappearance from the circulation.75,93 Se-rial small animal PET imaging with 64Cu-DOTA-minibodydemonstrated rapid and specific localization to the positivetumor with increasing accumulation of activity over time(from 2 to 24 hours p.i.) that was !20% ID/g at 24 hourswith a tumor-to-soft-tissue ratio of 6.93 As expected, highpersistent activity in the liver was also seen with activity movinginto the GI-tract at the late time-point. PET images with 124I-minibody produced excellent images at 18 hours p.i., with tu-mor uptakes of 21% ID/g and tumor-to-background ratio of26.2.75 The high ratio is due to the rapid clearance from theblood combined with fast metabolism and clearance of radioio-dine from nontargeted tissues. In comparison with FDG scan,

both radiometal labeled and radioiodinated minibody produceda much cleaner and higher contrast image.

Minibodies have also been produced against HER2 andPSCA expressed in breast and prostate cancers.94-96 Althoughboth anti-HER2 and anti PSCA minibodies demonstratedsimilar pharmacokinetics to the anti-CEA minibody and lo-calized specifically to tumors, the accumulation in tumorswas considerable lower. Activity in the tumor was about 5%ID/g at 12-24 hours with anti-HER2 131I-10H8 minibodywith tumor-to-blood ratios of 1.1 at 12 hours that increasedto 3 at 48 hours.96 This lower activity was explained by in-ternalization of the minibody on binding, followed by me-tabolism and rapid excretion of radiolabel from the cells.Biodistribution with anti-HER2 111In-DOTA-10H.8 mini-

T84.66 Mb

Rx. Mb

21 h

2B3 Mb

21 h

18 h

Herc. Mb

C6.5 Mb

20 h

Herc. Mb

20 h

Rx. Mb

19 h

T84.66 Mb

18 h

12 h

64Cu

124I

CEA CD20 HER2 PSCA

Figure 4 Small animal PET imaging of xenograft-bearing mice injected with 124I-labeled (top) and 64Cu-DOTA mini-bodies (bottom) against different cell surface antigens. Minibodies against CEA (T84.66 Mb),75 CD20 (Rx. Mb),99 HER2(Herc. Mb)99 were radiolabeled and imaged with both radionuclides. The anti-HER2 C6.5Mb and anti-PSCA 2B3 Mb94

were radiolabled with 124I only. All 124I-labeled minibodies, except the 124I-Herc. Mb, produce high-contrast images at18-21 hours p.i. Less contrast is achieved with 64Cu-DOTA minibodies due to high-background activity. Hepaticclearance is evident with the 64Cu-DOTA minibodies as expected for a fragment of !60 kDa in size.


Figure 4 Small animal PET imaging of xenograft-bearing mice injected with 124I-labeled (top) and 64Cu-DOTA minibodies (bottom) against different cell surface antigens. Minibodies against CEA (T84.66 Mb),75 CD20 (Rx. Mb),99 HER2 (Herc. Mb)99 were radiolabeled and imaged with both radionuclides. The anti-HER2 C6.5Mb and anti-PSCA 2B3 Mb94 were radiolabled with 124I only. All 124I-labeled minibodies, except the 124I-Herc. Mb, produce high-contrast images at 18-21 hours p.i. Less contrast is achieved with 64Cu-DOTA minibodies due to high-background activity. Hepatic clearance is evident with the 64Cu-DOTA minibodies as expected for a fragment of >60 kDa in size.


Minibody






21 h

64Cu 124I

18 h64Cu 124I

20 h18 h

A

B









21 h

64Cu 124I

18 h64Cu 124I

20 h18 h

A

B









21 h

64Cu 124I

18 h64Cu 124I

20 h18 h

A

B




scFv-Fc double mutant does not bind the nFcR

scFv-Fc



• There are several factors, such as the molecular weight, Fc domains, valency, and specificity that will influence targeting and pharmacokinetics

• Fragments above 60 kDa will clear through the liver, whereas those below this size will clear through the kidneys

• Presence of Fc domains will increase the serum residence time, and increase the exposure time to the target

• However, for imaging purposes, a more rapid clearance from the blood is warranted to increase contrast and sensitivity

• Difficulties in localizing and accumulating activity in target tissue are related to specificity and avidity

• In addition, the number of target antigen per cell and whether it internalizes or not will affect accumulation in target tissue

Factors that influence recombinant Ab pharmacokinetics

Documents

Antibody engineering Part 5 - unige.it · of antibody and antibody fragments including bacteria, yeast, plants, insect cells, and mammalian cells • Each has advantages, potential